Task allocation system and method for collaborative operation of deep-sea unmanned surface vehicle and underwater robot

Abstract

The invention provides a task allocation system and method for collaborative operation of a deep-sea unmanned ship and an underwater robot. The task allocation system comprises an environment sensing unit, a digital twin unit, a task modeling unit, a collaborative allocation unit, a collaborative scheduling unit, a dynamic reconstruction unit, a cross-domain management unit and a man-machine collaborative intervention unit. The environment sensing unit is used for collecting multi-source heterogeneous sensing data and semantic modeling of a dynamic ocean field; the digital twin unit is used for body state mirroring and capability degradation prediction; compared with the prior art, the method has the following beneficial effects that deep fusion and prediction of a global dynamic environment and a platform state are realized by constructing a cross-domain environment perception and digital twinborn model, a semantic task graph generation and distributed collaborative decision-making mechanism is utilized, a task scheduling capability of communication perception and a rapid reconstruction mechanism under an abnormal condition are also provided, and the method has a wide application prospect. Cooperative intelligence of multiple platforms in a complex deep sea environment is improved, and robustness and adaptability of task execution are remarkably enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of cooperative operation technology, in particular to a task allocation system and method for cooperative operation of a deep-sea unmanned surface vehicle and an underwater robot. BACKGROUND

[0002] A deep-sea unmanned surface vehicle is an unmanned intelligent platform that mainly navigates on the water surface and can perform autonomous operation, usually equipped with various communication, navigation and detection devices, and undertakes tasks such as water surface monitoring, communication relay and support for underwater equipment. An underwater robot mainly refers to an unmanned underwater vehicle that can perform various operation tasks autonomously or remotely in underwater environment, including autonomous underwater vehicles and remotely operated underwater robots, which usually carry various underwater sensors and operation tools for deep-sea exploration, scientific investigation and engineering operation tasks. Generally, the cooperative operation of the deep-sea unmanned surface vehicle and the underwater robot mainly relies on pre-programmed scripts and centralized control mode. During the operation process, the unmanned surface vehicle acts as a communication bridge between the mother ship and the underwater robot, and the task content and execution path are mostly set in advance before the operation. The underwater robot basically follows the established program during the task execution, and the unmanned surface vehicle is responsible for monitoring and limited data transfer.

[0003] Therefore, in the traditional task allocation method, a static allocation and manual intervention combined strategy is usually adopted. The control center allocates tasks to each platform according to experience, and the coordination between tasks and the real-time adjustment capability are weak. Due to the limited underwater communication bandwidth and high delay, the traditional method is difficult to perform dynamic task reallocation according to real-time environmental changes and platform states. Once equipment failure, environmental disturbance or sudden task occurs, the response of the whole system lags behind, and in serious cases, it even leads to task interruption. In addition, the traditional architecture lacks overall benefit optimization and resource balanced utilization mechanism for multi-platform cooperation, and the task execution is not flexible. SUMMARY

[0004] In view of the above-mentioned shortcomings of the prior art, the purpose of the present application is to provide a task allocation system and method for cooperative operation of a deep-sea unmanned surface vehicle and an underwater robot, which solves the problems that the traditional method is difficult to perform dynamic task reallocation according to real-time environmental changes and platform states, and lacks overall benefit optimization and resource balanced utilization mechanism for multi-platform cooperation, and the task execution is not flexible.

[0005] To achieve the above-mentioned purpose and other related purposes, the present application provides the following technical scheme:

[0006] The task allocation system for the deep-sea unmanned vehicle and underwater robot cooperative operation comprises an environment perception unit, a digital twin unit, a task modeling unit, a cooperative allocation unit, a cooperative scheduling unit, a dynamic reconstruction unit, a cross-domain management unit and a man-machine cooperative intervention unit; the environment perception unit is used for collecting multi-source heterogeneous sensing data and semantic modeling of dynamic ocean field; the digital twin unit is used for ontology state mirroring and capability degradation prediction; the task modeling unit is used for analyzing mission instructions and generating a semantic task graph; the cooperative allocation unit is used for decentralized task negotiation;

[0007] The cooperative scheduling unit is used for communication awareness and information distribution; the dynamic reconstruction unit is used for abnormal detection analysis and dynamic negotiation of task contracts; the cross-domain management unit is used for energy state consensus and scheduling autonomous energy; the man-machine cooperative intervention unit is used for multi-dimensional situation visualization, AR enhancement and injection of high-order instructions.

[0008] In an embodiment of the present application, the environment perception unit comprises a multi-source heterogeneous sensing data acquisition module and a dynamic ocean field semantic modeling module; the multi-source heterogeneous sensing data acquisition module is used for acquiring full-dimensional environment raw data; the dynamic ocean field semantic modeling module is used for constructing a dynamic four-dimensional environment map containing a sea current field, an obstacle probability distribution and a communication quality field semantic information according to the acquired full-dimensional environment raw data.

[0009] In an embodiment of the present application, the digital twin unit comprises an ontology state mirroring module and a capability degradation prediction module; the ontology state mirroring module is used for constructing millisecond-level synchronization and mapping of the physical state of the physical platform in the digital space; the capability degradation prediction module is used for predicting the degradation curves of the positioning accuracy, speed range and task execution capability of the physical platform at any future time.

[0010] In an embodiment of the present application, the task modeling unit comprises a mission instruction analysis module and a semantic task graph generation module; the mission instruction analysis module is used for analyzing core operation verbs, geographical objects and task attributes in the mission instruction; the semantic task graph generation module is used for converting the abstract mission instruction into a structured and calculable task network.

[0011] In an embodiment of the present application, the cooperative allocation unit comprises a task utility calculation module and a distributed consensus module; the task utility calculation module is used for interacting with the digital twin of the physical platform, so as to simulate the utility value of each physical platform when executing each sub-task; the distributed consensus module is used for enabling the heterogeneous physical platform group to quickly reach Nash equilibrium on the task allocation scheme.

[0012] In an embodiment of the present application, the cooperative scheduling unit comprises a communication awareness module and an information distribution module; the communication awareness module is used to plan a motion trajectory which can not only complete its own task but also keep in a reliable communication window with the relay USV; the information distribution module is used to determine when, where and to whom to relay or piggyback what kind of data.

[0013] In an embodiment of the present application, the dynamic reconstruction unit comprises an anomaly detection analysis module and a task contract dynamic negotiation module; the anomaly detection analysis module is used to diagnose single-point failures and deduce their cascading effects on the entire task network; the task contract dynamic negotiation module is used to smooth redistribution of system tasks and resilience continuation of the overall task.

[0014] In an embodiment of the present application, the cross-domain management unit comprises an energy state consensus module and an autonomous energy scheduling module; the energy state consensus module is used to establish a shared and trusted global energy state view at the group level; the autonomous energy scheduling module is used to guide the physical platform to perform safe energy cooperative action in the extreme case of communication interruption.

[0015] In an embodiment of the present application, the human-machine cooperative intervention unit comprises a multi-dimensional situation visualization module and a high-level instruction injection module; the multi-dimensional situation visualization module is used to provide intuitive and immersive global situation awareness; the high-level instruction injection module is used to compile it into a semantic task graph modification instruction that the system can understand, and inject it into task decomposition and distribution.

[0016] A task allocation method for deep-sea unmanned vehicle and underwater robot cooperative operation, based on the task allocation system for deep-sea unmanned vehicle and underwater robot cooperative operation, comprising the following steps: acquiring multi-physical field data synchronously collected by each physical platform, constructing a four-dimensional dynamic environment map according to the multi-physical field data, and distributing the four-dimensional dynamic environment map to all physical platforms participating in cooperative operation;

[0017] Acquiring key parameters related to itself monitored by each physical platform in real time, and predicting the performance decay curve of each physical platform at any future time based on the historical task load data and the currently acquired key parameters of each physical platform; receiving high-level task instructions issued by the operator and analyzing them, then converting the abstract mission parsed into a semantic task graph, and performing instantiation processing on the semantic task graph according to the four-dimensional dynamic environment map and the performance decay curve;

[0018] The specific task item generated by the instantiation processing is sent to all physical platforms, each physical platform simulates the utility value required for executing each task, and then the physical platforms quickly reach a Nash equilibrium on the task allocation scheme; each physical platform plans a local motion trajectory according to the task contract allocated, and predicts the future communication opportunity window of itself and other physical platforms according to the predicted motion trajectory and the communication quality map in the four-dimensional dynamic map;

[0019] Key state data monitored by each physical platform during task execution is acquired, and abnormality detection is performed on the key state data, and smooth reallocation of system tasks and resilience continuation of overall tasks are realized according to the abnormality detection result; key information of the virtual physical platforms is superimposed into real ocean scene video streams, and displayed on a monitoring interface, and then new policy instructions issued by an operator under global situation awareness are received.

[0020] As described above, the task allocation system and method for deep-sea unmanned vehicles and underwater robots to cooperate in the present application have the following beneficial effects: the present application realizes deep fusion and prediction of global dynamic environment and platform state by constructing cross-domain environment perception and digital twin model, and has task scheduling ability and rapid reconstruction mechanism in abnormal conditions with the help of semantic task graph generation and distributed collaborative decision mechanism, which not only improves the collaborative intelligence of multiple platforms in complex deep-sea environment, but also significantly enhances the robustness and adaptability of task execution. At the same time, the system of the present application supports human-machine collaborative intervention, and the operator can issue high-level policy instructions to further improve the flexibility and reliability of the overall system. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The overall topology diagram of the task allocation system for deep-sea unmanned vehicles and underwater robots to cooperate in the embodiments of the present application is shown;

[0022] Figure 2 The overall flowchart of the task allocation method for deep-sea unmanned vehicles and underwater robots to cooperate in the embodiments of the present application is shown. DETAILED DESCRIPTION

[0023] The embodiments of the present application will be described below through specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the disclosure of the present specification. It should be noted that the following examples and features in the examples can be combined with each other without conflict.

[0024] The first embodiment of the present application relates to a task allocation system for deep-sea unmanned vehicles and underwater robots to cooperate, please refer to Figure 1, including an environment perception unit, a digital twin unit, a task modeling unit, a collaborative allocation unit, a collaborative scheduling unit, a dynamic reconfiguration unit, a cross-domain management unit, and a human-machine collaborative intervention unit; the environment perception unit is configured to collect multi-source heterogeneous sensing data and semantic modeling of dynamic ocean fields; the digital twin unit is configured to perform ontology state mirroring and capability degradation prediction; the task modeling unit is configured to analyze mission instructions and generate a semantic task graph; the collaborative allocation unit is configured to perform decentralized task negotiation; the collaborative scheduling unit is configured to perform communication awareness and information distribution; the dynamic reconfiguration unit is configured to perform abnormality detection analysis and dynamic negotiation of task contracts; the cross-domain management unit is configured to perform energy state consensus and scheduling of autonomous energy; and the human-machine collaborative intervention unit is configured to perform multi-dimensional situation visualization, AR enhancement, and injection of high-level instructions.

[0025] The environment perception unit includes a multi-source heterogeneous sensing data acquisition module and a dynamic ocean field semantic modeling module; the multi-source heterogeneous sensing data acquisition module acquires electromagnetic, acoustic, and ocean physical field data through USV-mounted Starlink communication, millimeter wave radar, AUV-mounted acoustic Doppler current profiler, and deep water hydrophone array, to realize acquisition of full-dimensional environment raw data; and the dynamic ocean field semantic modeling module fuses and reasons sparse and asynchronous data in the acquired multi-physical field data through a graph neural network, to construct a dynamic four-dimensional environment map containing semantic information of a sea current field, an obstacle probability distribution, and a communication quality field.

[0026] The digital twin unit includes an ontology state mirroring module and a capability degradation prediction module; the ontology state mirroring module integrates real-time telemetry data of the physical platform, such as power, cabin temperature, thruster efficiency, and device health state, and drives a high-precision dynamics model, to realize millisecond-level synchronization and mapping of the physical state of the physical platform in the digital space; and the capability degradation prediction module analyzes the relationship between historical task load and performance degradation through a time series-based LSTM deep learning method, to predict the degradation curve of the positioning accuracy, speed range, and task execution capability of the physical platform at any future time, to provide forward-looking constraints for task allocation.

[0027] The task modeling unit includes a mission instruction analysis module and a semantic task graph generation module; the mission instruction analysis module analyzes the high-level command of “thoroughly survey the east slope of Seamount A and identify suspected cold-water coral areas” through NLP natural language processing, to extract the core operation verbs, the Seamount A, the east slope geographical object, and the task attributes; and the semantic task graph generation module constructs the extracted elements into a semantic graph containing task nodes, including “multi-beam scanning” and “near-optical identification”, node relationship such as time sequence, dependency, and replacement, and task requirement metadata such as required sensors, accuracy, and priority, to convert the abstract mission into a structured and computable task network.

[0028] The cooperative allocation unit comprises a task utility calculation module and a distributed consensus module; the task utility calculation module interacts with the platform digital twin through a partially observable Markov decision process, simulates future energy consumption, time, risk and cost of performing a task, and is used for calculating the "utility evaluation" of each physical platform for each subtask and serving as a basis for bidding; the distributed consensus module performs multi-round iterative games between the unmanned surface vehicle and the underwater robot group under limited communication through an improved alternating direction multiplier method, so that the heterogeneous physical platform group quickly reaches a Nash equilibrium on the task allocation scheme, and forms a distributed protocol that meets the global benefit and respects individual rationality.

[0029] The cooperative scheduling unit comprises a communication awareness module and an information distribution module; the communication awareness module embeds the channel capacity and delay prediction model of the underwater acoustic communication link as a constraint condition into the model predictive control algorithm, so as to plan a motion trajectory that can complete the task itself and best keep the reliable communication window with the relay USV; the information distribution module pre-calculates the communication opportunity window on the future trajectory of each physical platform, constructs a "delay tolerant network" routing table, and decides when, where and to whom to relay or "piggyback" what kind of data, including state information, emergency alarm and non-real-time image, so as to maximize the use of each short communication opportunity.

[0030] The dynamic reconfiguration unit comprises an anomaly detection analysis module and a task contract dynamic negotiation module; the anomaly detection analysis module compares the digital twin predicted state with the actual state in real time, and uses the Bayesian network to model the functional dependence between platforms, so as to quickly diagnose single point failures and deduce their cascading effects on the entire task network; the task contract dynamic negotiation module triggers a Shapley value-based coalition game mechanism to re-auction the task contracts released or not completed due to failures, and fairly calculates compensation according to the remaining capacity of each physical platform, so as to realize smooth redistribution of system tasks and resilience continuation of overall tasks.

[0031] The cross-domain management unit comprises an energy state consensus module and an autonomous energy scheduling module; the energy state consensus module synchronizes the remaining power and energy consumption rate information between platforms through a lightweight consensus protocol, so as to establish a shared and trusted global energy state view at the group level; the autonomous energy scheduling module uses pre-defined rules, such as "when the USV power is higher than 80% and a certain AUV power is detected to be lower than 20%, the USV actively approaches it", to guide the platform to perform safe energy coordination actions in the case of communication interruption, thereby prolonging the overall operation time of the system.

[0032] The human-machine collaborative intervention unit includes a multi-dimensional situation visualization module and a high-order command injection module. The multi-dimensional situation visualization module renders the state of the cross-domain digital twin in real time through a game engine, and overlays key information such as predicted trajectory, communication link, and task progress onto the real ocean scene video stream in an AR manner, providing operators with an intuitive and immersive global situational awareness. The high-order command injection module receives strategy-level commands issued by operators through voice or graphical interfaces, including "prioritize the reconnaissance of area B" and "AUV formation immediately avoid strong current areas," and compiles them into semantic task graph modification commands that the system can understand, and injects them into task decomposition and allocation, realizing the integration of experts and artificial intelligence.

[0033] The second embodiment of the present invention relates to a task allocation method for collaborative operation between a deep-sea unmanned surface vessel and an underwater robot, the process of which is as follows: Figure 2 As shown, the details are as follows:

[0034] Step 101: Obtain multi-physics field data synchronously collected by each physical platform, construct a four-dimensional dynamic environment map based on the multi-physics field data, and distribute the four-dimensional dynamic environment map to all physical platforms participating in the collaborative operation.

[0035] Specifically, step 101 can also be summarized as environmental data acquisition and global situational awareness construction, which includes the following steps:

[0036] 1.1: Multi-physics field data synchronous acquisition: The unmanned surface vessel acquires macro-oceanic meteorological forecast data through its satellite communication link, and at the same time uses millimeter-wave radar to scan for obstacles on the water surface; each underwater robot uses its onboard acoustic sensor group to continuously collect hydrological data such as ocean current speed, direction, and water temperature at its location, as well as underwater acoustic communication channel status data.

[0037] 1.2: Dynamic Environment Model Fusion Calculation: The sparse and asynchronous multi-source data collected above are spatiotemporally aligned and input into a trained neural network model; the model outputs a four-dimensional dynamic map with probabilistic characteristics covering the entire operating area. The map clearly marks the ocean current vectors at different locations, known and suspected obstacle areas, and the expected underwater acoustic communication quality between different locations.

[0038] 1.3: Global Situation Sharing: The generated dynamic environment map serves as the common basis for all subsequent decisions. It is distributed to all unmanned surface vessels and underwater robot platforms participating in the collaborative operation through reliable communication links, using radio on the surface and underwater acoustic communication windows.

[0039] Step 102, obtain the key parameters related to itself monitored by each physical platform in real time, and predict the performance degradation curve of each physical platform at any future time based on the historical task load data and the currently obtained key parameters of each physical platform.

[0040] Specifically, step 102 can also be briefly described as platform real-time state synchronization and performance prediction, which specifically includes the following steps:

[0041] 2.1: Platform state reporting and mirroring: Each physical platform continuously monitors its own power, device temperature, thruster speed, task load health status and other key parameters, and packages these data into a state message; When establishing a communication connection with the command node each time, including the unmanned ship, the state message is preferentially reported to update a high-fidelity platform dynamics virtual model running in the control center;

[0042] 2.2: Performance degradation prediction: Based on the historical task load data and the currently reported health status of the physical platform, the time series prediction algorithm is used to predict the key performance indicators of each platform in the future period of time: maximum speed, positioning accuracy, sensor effective action distance, and generate its performance degradation curve;

[0043] 2.3: Generate capability constraint boundary: Convert the predicted performance degradation curve into the capability constraint boundary of the physical platform when executing tasks in the future: "one hour later, the maximum cruising speed of the AUV is reduced to 2 knots", this boundary data will be directly used to constrain the subsequent task allocation and path planning algorithm.

[0044] Step 103, receive the high-level task instruction issued by the operator, analyze it, and then convert the abstract mission parsed out into a semantic task graph, and perform instantiation processing on the semantic task graph according to the four-dimensional dynamic environment map and the performance degradation curve.

[0045] Specifically, step 103 can also be briefly described as high-level instruction analysis and executable task generation, which specifically includes the following steps:

[0046] 3.1: Natural language instruction structured analysis: Receive the high-level task instruction issued by the operator: for example: "conduct detailed investigation on the north slope of sea mountain B, and mark all suspected mineral outcrops as the focus"; Through natural language processing technology, the core actions "detailed investigation", "marking" in the instruction, the geographical area "north slope of sea mountain B" and the task attributes "focus", "suspected" are identified;

[0047] 3.2: Constructing semantic task graph: According to the parsed structured information, call the corresponding task template from a predefined task knowledge base, generate a directed graph composed of multiple task nodes, the nodes in the graph represent specific atomic tasks: "execute 200-meter line-scan sonar survey", "perform spectral sampling on coordinate (X, Y)", the edges between nodes represent the logical relationship between tasks, including execution order, dependency relationship, task B must start after task A is completed or replacement relationship;

[0048] 3.3: Injecting environmental and platform constraints: Take the dynamic environmental map generated by S101 and the platform capability constraint boundary generated by S102 as input conditions, "instance" the semantic task graph; for example, calculate the specific scan route for the "line-scan sonar survey" task node that avoids strong flow areas; or specify a platform with sufficient power and healthy mechanical arm performance for the "spectral sampling" task.

[0049] Step 104, send the specific task items generated by the instantiation process to all physical platforms, each physical platform simulates the utility value required to execute each task, and then makes each physical platform quickly reach a Nash equilibrium on the task allocation scheme.

[0050] Specifically, step 104 can also be described as distributed task negotiation and allocation, which includes the following steps:

[0051] 4.1: Task bidding based on utility: "broadcast" the specific task items generated by S103 to all available platforms; after each physical platform receives the task list, it combines its S102 state and S101 environmental information to simulate the cost, time, energy consumption and benefit required to execute each task, and calculates a "utility value" as a bid price for the task;

[0052] 4.2: Multi-round iterative consensus: Each physical platform sends the bid price to the command node or exchanges it among platforms through the underwater acoustic network; use a distributed optimization algorithm for multi-round iterative calculation, and finally make all physical platforms reach a consensus scheme on "which task is executed by which platform", which pursues the highest efficiency as a whole while respecting the individual ability limits of each platform;

[0053] 4.3: Task contract signing: After the consensus scheme is formed, the command node distributes self-consistent and detailed task instructions to each platform, which is equivalent to signing a "task contract", and the contract content includes specific target points, paths, actions and success criteria.

[0054] Step 105, each physical platform plans the local motion trajectory according to the assigned task contract, and predicts the future communication opportunity window of itself and other physical platforms according to the predicted motion trajectory and the communication quality map in the four-dimensional dynamic map.

[0055] Specifically, step 105 can also be briefly described as communication-aware collaborative execution and dynamic scheduling, which specifically includes the following steps:

[0056] 5.1: Trajectory and communication joint planning: Each physical platform plans a local motion trajectory according to the assigned task contract. In the planning, not only obstacle avoidance and energy consumption are considered, but also the requirement of maintaining reliable underwater acoustic communication with the relay unmanned ship is taken as one of the core constraints to ensure that the platform is in the communication window at the key nodes.

[0057] 5.2: Predictive opportunistic communication: Each physical platform predicts its future communication opportunity window with other physical platforms or unmanned ships based on the predicted motion trajectory and the communication quality map in S101; according to the priority of data: state data > alarm data > non-real-time image data, a data distribution plan is generated to determine when to send which type of data, or whether to "piggyback" data for neighboring platforms.

[0058] 5.3: Closed-loop state monitoring and execution: Each physical platform starts to execute the task and continuously sends back the key state data as planned. The command node monitors the progress of the entire task execution to ensure that the system runs as planned.

[0059] Step 106, obtain the key state data monitored by each physical platform during task execution, and perform anomaly detection on it, and realize smooth reallocation of system tasks and resilience continuation of overall tasks according to the anomaly detection results.

[0060] Specifically, step 106 can also be briefly described as anomaly diagnosis and system resilience recovery, which specifically includes the following steps:

[0061] 6.1: Anomaly detection and root cause analysis: By comparing the actual returned state data of the platform with the predicted state of the virtual model in S102, anomalies are detected in real time. If the platform is lost or the task progress is far behind the plan, the potential impact of the anomaly on the overall task network is analyzed using the dependency graph.

[0062] 6.2: Task contract re-auction and compensation: If it is confirmed that a physical platform fails to complete the task, the uncompleted task contract is marked as "to be reallocated", triggering a fast renegotiation process to re-auction the to-be-allocated task to other available platforms, and according to the new bidding scheme, a fair compensation calculation is made for the platforms that have increased their burden due to taking on new tasks, and priority is given in subsequent task allocation.

[0063] 6.3: System reconstruction and continued execution: The new allocation scheme is issued to the relevant platforms, and each physical platform adjusts its original plan to seamlessly access and execute the newly allocated tasks, thereby ensuring that the overall mission can continue to be executed to the maximum extent even in the case of individual failure.

[0064] Step 107, superimpose the key information of each virtual physical platform into the real marine scene video stream, and display on the monitoring interface, then receive the new strategy instruction issued by the operator under the global situation awareness.

[0065] Specifically, step 107 can also be briefly described as human-in-the-loop supervision and strategy intervention, which specifically includes the following steps:

[0066] 7.1: Immersive situation monitoring: The operator's monitoring interface does not display raw data, but presents an augmented reality view based on the data fusion of S101, S102 and S103. The operator can visually see the superposition of virtual physical platform icons on the real marine scene video, their predicted trajectories, communication links and task completion degree;

[0067] 7.2: High-level strategy instruction issuance: Based on the global situation, if the operator judges that the automatic system decision needs to be adjusted, he / she can issue a strategy-level instruction, "all physical platforms immediately suspend work and rise to 50 meters to avoid the strong internal wave that will pass soon", which is received and understood by the system;

[0068] 7.3: System instruction fusion and re-planning: The system converts the operator's strategy instruction into a modification of the current S103 semantic task graph, adds a parallel pre-task of "rising to 50 meters" to all task nodes, then automatically jumps to S104 to re-perform fast distributed negotiation and allocation based on the modified task graph, generates a new execution scheme and issues it, so that the whole system responds to the decision smoothly.

[0069] In summary, the present application realizes the deep fusion and prediction of global dynamic environment and platform state, with the help of semantic task graph generation and distributed collaborative decision-making mechanism, it also has the ability of communication-aware task scheduling and fast reconfiguration mechanism in abnormal situations, which not only improves the collaborative intelligence of multiple platforms in complex deep sea environment, but also significantly enhances the robustness and adaptability of task execution. At the same time, the system of the present application supports human-machine collaborative intervention, the operator can intervene in the control loop and issue high-level strategy instructions, further improving the flexibility and reliability of the overall system.

[0070] The above embodiments only exemplarily illustrate the principles and effects of the present application, and are not used to limit the present application. All equivalent modifications or changes made by those skilled in the art without departing from the spirit and technical thought disclosed by the present application should be covered by the claims of the present application.

Claims

1. A task allocation system for collaborative operation between a deep-sea unmanned surface vessel and an underwater robot, characterized in that: It includes an environmental perception unit, a digital twin unit, a task modeling unit, a collaborative allocation unit, a collaborative scheduling unit, a dynamic reconfiguration unit, a cross-domain management unit, and a human-machine collaborative intervention unit; The environmental perception unit is used to collect multi-source heterogeneous sensor data and semantic modeling of dynamic ocean fields; the digital twin unit is used for ontology state mirroring and capability degradation prediction; the task modeling unit is used to parse mission instructions and generate semantic task graphs; and the collaborative allocation unit is used for decentralized task negotiation. The collaborative scheduling unit is used for communication sensing and information distribution; the dynamic reconstruction unit is used for anomaly detection and analysis and dynamic negotiation of task contracts; the cross-domain management unit is used for energy status consensus and scheduling autonomous energy; and the human-machine collaborative intervention unit is used for multi-dimensional situation visualization and AR enhancement and injection of high-order instructions.

2. The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots according to claim 1, characterized in that: The environmental perception unit includes a multi-source heterogeneous sensor data acquisition module and a dynamic ocean field semantic modeling module. The multi-source heterogeneous sensor data acquisition module is used to acquire raw environmental data in all dimensions; the dynamic ocean field semantic modeling module is used to construct a dynamic four-dimensional environmental map containing semantic information of ocean current field, obstacle probability distribution, and communication quality field based on the acquired raw environmental data in all dimensions.

3. The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots according to claim 1, characterized in that: The digital twin unit includes an ontology state mirroring module and a capability degradation prediction module; The physical state mirroring module is used to construct millisecond-level synchronization and mapping of the physical state of the physical platform in the digital space; the capability degradation prediction module is used to predict the degradation curve of the positioning accuracy, speed range and task execution capability of the physical platform at any future time.

4. The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots according to claim 1, characterized in that: The task modeling unit includes a mission instruction parsing module and a semantic task graph generation module; The mission instruction parsing module is used to parse out the core operation verbs, geographic objects, and task attributes in the mission instructions; the semantic task graph generation module is used to transform the parsed abstract mission instructions into a structured and computable task network.

5. The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots according to claim 1, characterized in that: The collaborative allocation unit includes a task utility calculation module and a distributed consensus module; The task utility calculation module is used to interact with the digital twin of the physical platform to simulate the utility value of each physical platform when executing each sub-task; the distributed consensus module is used to enable the heterogeneous physical platform group to quickly reach Nash equilibrium on the task allocation scheme.

6. The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots according to claim 1, characterized in that: The collaborative scheduling unit includes a communication sensing module and an information distribution module; The communication sensing module is used to plan a movement trajectory that can both complete its own task and maintain optimal communication with the relay USV within a reliable communication window; the information distribution module is used to determine when, where, to whom, or what type of data to relay or carry.

7. The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots according to claim 1, characterized in that: The dynamic reconstruction unit includes an anomaly detection and analysis module and a task contract dynamic negotiation module; The anomaly detection and analysis module is used to diagnose single-point failures and deduce their cascading impact on the entire task network; the task contract dynamic negotiation module is used for the smooth redistribution of system tasks and the resilient continuation of overall tasks.

8. The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots according to claim 1, characterized in that: The cross-domain management unit includes an energy status consensus module and an autonomous energy scheduling module; The energy state consensus module is used to establish a shared and trusted global energy state view at the group level; the autonomous energy scheduling module is used to guide the physical platform to perform energy coordination actions to ensure security in extreme cases of communication interruption.

9. The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots according to claim 1, characterized in that: The human-machine collaborative intervention unit includes a multi-dimensional situation visualization module and a high-order instruction injection module; The multi-dimensional situation visualization module is used to provide intuitive and immersive global situation awareness; the high-order instruction injection module is used to compile it into semantic task graph modification instructions that the system can understand, and inject it into task decomposition and allocation.

10. A task allocation method for collaborative operations between a deep-sea unmanned surface vessel and an underwater robot, characterized in that: The task allocation system for collaborative operation of deep-sea unmanned surface vessels and underwater robots based on any one of claims 1 to 9 includes the following steps: Acquire multiphysics field data synchronously collected from each physical platform, construct a four-dimensional dynamic environment map based on the multiphysics field data, and distribute the four-dimensional dynamic environment map to all physical platforms participating in the collaborative operation; Obtain key parameters related to each physical platform in real time, and predict the performance degradation curve of each physical platform at any future time based on the historical task load data of each physical platform and the currently obtained key parameters. The system receives advanced task instructions from the operator, parses them, transforms the parsed abstract task into a semantic task graph, and instantiates the semantic task graph based on the four-dimensional dynamic environment map and performance degradation curve. The instantiated task items generated through instantiation are sent to all physical platforms. Each physical platform simulates the utility value required to execute each task, and then the physical platforms quickly reach a Nash equilibrium on the task allocation scheme. Each physical platform plans its local motion trajectory according to the assigned task contract, and predicts its future communication opportunity window with other physical platforms based on the predicted motion trajectory and the communication quality map in the four-dimensional dynamic map. Acquire key status data monitored by each physical platform during task execution, perform anomaly detection, and achieve smooth redistribution of system tasks and resilient continuation of overall tasks based on the anomaly detection results. The key information of each virtual physical platform is superimposed onto the video stream of the real ocean scene and displayed on the monitoring interface. Then, new strategy instructions issued by the operator under global situational awareness are received.

Images