A decision-making system for underwater unmanned platforms based on fast and slow brain routing arbitration and related products
By introducing a fast-slow brain routing arbitration mechanism into the underwater unmanned platform decision-making system, a comprehensive judgment and safety verification of candidate solutions are performed, solving the problem of lack of safety and feasibility judgment in the existing technology, and improving the safety and reliability of underwater unmanned platform decision-making.
Patent Information
- Application Number
- CN202611351520.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-02
- Publication Date
- 2026-11-17
AI Technical Summary
In existing technologies, large language models lack security and feasibility assessment after generating decision-making schemes for underwater unmanned platforms. Especially under weak communication conditions, this leads to unexecutable decision-making schemes being directly sent to the underwater unmanned platform for execution, affecting the safety and reliability of the mission.
A fast and slow brain routing arbitration mechanism is introduced. By setting up a router module between the slow brain module and the fast brain module, the candidate solutions generated by the slow brain module are comprehensively judged and security verified. Only the solutions that pass the verification can be sent to the fast brain module to generate real-time control decisions, thus building an effective verification and interception mechanism.
This effectively avoids the direct issuance of unfeasible solutions, ensuring the safety and reliability of underwater unmanned platform decision-making. By completing the safety and feasibility assessment before issuing and executing candidate solutions, the safety and reliability of the mission are improved.
Smart Images

Figure CN123115337A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to an underwater unmanned platform decision-making system and related products based on fast and slow brain routing arbitration. Background Technology
[0002] In recent years, with the rapid development of artificial intelligence technology, the application of large language models in the field of underwater unmanned platform decision-making has attracted increasing attention. Large language models, with their advantages in semantic understanding, knowledge reasoning, and complex scene cognition, provide a new technical path for building underwater autonomous decision-making systems with high-level situational awareness and mission planning capabilities.
[0003] However, in existing technologies, decision-making schemes generated by incorporating large language models into underwater unmanned platforms are often directly deployed to the platform for execution, lacking a step to assess the safety and feasibility of the decision-making schemes themselves. Especially under weak communication conditions, large language models may generate unexecutable decision-making schemes based on outdated or incomplete situational information. The lack of a pre-verification mechanism will result in such unexecutable decision-making schemes being directly deployed to the underwater unmanned platform, seriously affecting the safety and reliability of the mission. Summary of the Invention
[0004] This application provides an underwater unmanned platform decision-making system and related products based on fast and slow brain routing arbitration. By setting up a router module between the slow brain module and the fast brain module, the candidate solutions generated by the slow brain module must undergo comprehensive judgment and security verification before entering the fast brain module. Only candidate solutions that pass the verification can be sent to the fast brain module, thus avoiding the security risks caused by unexecutable solutions being directly sent to the underwater unmanned platform for execution.
[0005] In a first aspect, embodiments of this application provide an underwater unmanned platform decision-making system based on fast and slow brain routing arbitration, the system including a slow brain module, a fast brain module and a router module; The slow brain module is used to acquire multi-source situational information, perform task reasoning on the multi-source situational information, generate candidate solutions, and output them to the router module. The router module is used to receive the candidate schemes, perform comprehensive judgment and security verification on the candidate schemes, and send the verified candidate schemes to the QuickBrain module. The fast brain module is used to receive the verified candidate solutions and generate real-time control decisions based on the verified candidate solutions.
[0006] In one feasible implementation, the fast brain module is further configured to send the real-time control decision to the underwater unmanned platform for execution, obtain the execution result of the underwater unmanned platform, and feed the execution result back to the router module; The router module is also used to receive the execution result, and when the execution result indicates that the execution failed, to notify the slow brain module to regenerate the candidate solution.
[0007] One feasible implementation includes the following steps: performing comprehensive judgment and security verification on the candidate solutions, and sending the verified candidate solutions to the QuickBrain module: Based on preset judgment conditions, the candidate solutions are comprehensively judged to obtain a judgment result; When the determination result is passed, the candidate solution is subjected to security verification to obtain the verification result. The security verification includes rule verification, permission verification, security boundary verification and resource constraint verification. When the verification result passes, the candidate solution is sent to the QuickBrain module.
[0008] One feasible implementation, wherein generating real-time control decisions based on the verified candidate schemes, includes: The candidate schemes that pass the verification are parsed to obtain the structured intent; Based on a preset behavior tree template, the structured intent is translated to obtain an action sequence; Based on a preset control algorithm, the action sequence is mapped to generate real-time control decisions.
[0009] One feasible implementation, wherein performing task reasoning on the multi-source situational information to generate candidate solutions, includes: Based on a preset task reasoning model, task reasoning is performed on the multi-source situational information to obtain the reasoning result; The reasoning results are packaged to generate the candidate solutions.
[0010] One feasible implementation method includes acquiring multi-source situational information, comprising: Receive multi-source input data, including task instruction data, platform sensor data, and communication link status data; The multi-source input data is standardized to obtain the multi-source situational information.
[0011] One feasible implementation method is that the preset judgment conditions include platform status conditions, communication quality conditions, and credibility scoring conditions. The platform status condition is used to determine whether the status of the underwater unmanned platform allows the execution of the candidate scheme; The communication quality conditions are used to determine whether the communication link meets the communication requirements required for the execution of the candidate scheme. The credibility scoring criteria are used to determine whether the credibility of the candidate solution meets the execution requirements.
[0012] Secondly, embodiments of this application provide a decision-making method for underwater unmanned platforms based on fast and slow brain routing arbitration, applied to the underwater unmanned platform decision-making system described in the first aspect, including: The slow brain module in the underwater unmanned platform decision-making system acquires multi-source situational information, performs task reasoning on the multi-source situational information, generates candidate solutions, and outputs them to the router module. The router module in the underwater unmanned platform decision-making system receives the candidate solutions, performs comprehensive judgment and security verification on the candidate solutions, and sends the verified candidate solutions to the QuickBrain module. The fast-brain module in the underwater unmanned platform decision-making system receives the verified candidate schemes and generates real-time control decisions based on the verified candidate schemes.
[0013] Thirdly, embodiments of this application provide an electronic device, the device including: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, which includes instructions that, when executed by the processor, cause the processor to perform the implementation steps of the underwater unmanned platform decision-making method based on fast and slow brain routing arbitration.
[0014] Fourthly, embodiments of this application provide a computer-readable storage medium for storing a computer program. When the computer program is executed by a terminal device, it implements the above-described decision-making method for underwater unmanned platforms based on fast and slow brain routing arbitration.
[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: As can be seen from the above technical solution, this application provides an underwater unmanned platform decision-making system based on fast and slow brain routing arbitration. The system includes a slow brain module, a fast brain module, and a router module. The slow brain module is used to acquire multi-source situational information, perform task reasoning on the multi-source situational information, generate candidate solutions, and output them to the router module. The router module is used to receive candidate solutions, perform comprehensive judgment and security verification on the candidate solutions, and send the verified candidate solutions to the fast brain module. The fast brain module is used to receive the verified candidate solutions and generate real-time control decisions based on the verified candidate solutions.
[0016] As can be seen, this application, by setting up a router module between the slow-brain module and the fast-brain module, ensures that candidate solutions generated by the slow-brain module undergo comprehensive judgment and security verification before entering the fast-brain module. Only candidate solutions that pass the verification can be sent to the fast-brain module, avoiding the security risks caused by unexecutable solutions being directly sent to the underwater unmanned platform for execution. Thus, the security and feasibility of the candidate solutions themselves are judged before being sent to the underwater unmanned platform for execution. Therefore, this application constructs an effective verification and interception mechanism before candidate solutions are sent for execution. Attached Figure Description
[0017] Figure 1 A schematic diagram of the structure of an underwater unmanned platform decision-making system based on fast and slow brain routing arbitration is provided for an embodiment of this application; Figure 2 A flowchart illustrating the switching of fast and slow brain dominance is provided in an embodiment of this application; Figure 3 A flowchart illustrating the candidate scheme degradation process under weak communication conditions is provided in an embodiment of this application. Figure 4 A flowchart of an underwater unmanned platform decision-making method based on fast and slow brain routing arbitration provided in this application embodiment; Figure 5 This is a structural block diagram of an underwater unmanned platform decision-making method based on fast and slow brain routing arbitration, provided in an embodiment of this application. Detailed Implementation
[0018] As mentioned earlier, in existing technologies, decision-making schemes generated by incorporating large language models into underwater unmanned platforms are often directly deployed to the platform for execution, lacking a step to assess the safety and feasibility of the decision-making schemes themselves. Especially under weak communication conditions, large language models may generate unexecutable decision-making schemes based on outdated or incomplete situational information. The lack of a pre-verification mechanism will result in such unexecutable decision-making schemes being directly deployed to the underwater unmanned platform, seriously affecting the safety and reliability of the mission.
[0019] This application provides an underwater unmanned platform decision-making system based on fast and slow brain routing arbitration. The system includes a slow brain module, a fast brain module, and a router module. The slow brain module acquires multi-source situational information, performs task reasoning on the multi-source situational information, generates candidate solutions, and outputs them to the router module. The router module receives candidate solutions, performs comprehensive judgment and security verification on the candidate solutions, and sends the verified candidate solutions to the fast brain module. The fast brain module receives the verified candidate solutions and generates real-time control decisions based on the verified candidate solutions.
[0020] As can be seen, this application, by setting up a router module between the slow-brain module and the fast-brain module, ensures that candidate solutions generated by the slow-brain module undergo comprehensive judgment and security verification before entering the fast-brain module. Only candidate solutions that pass the verification can be sent to the fast-brain module, avoiding the security risks caused by unexecutable solutions being directly sent to the underwater unmanned platform for execution. Thus, the security and feasibility of the candidate solutions themselves are judged before being sent to the underwater unmanned platform for execution. Therefore, this application constructs an effective verification and interception mechanism before candidate solutions are sent for execution.
[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0022] Figure 1 This is a schematic diagram of the structure of an underwater unmanned platform decision-making system based on fast and slow brain routing arbitration, provided as an embodiment of this application. (Combined with...) Figure 1 As shown, the underwater unmanned platform decision-making system based on fast and slow brain routing arbitration includes a slow brain module, a router module, and a fast brain module.
[0023] Specifically, the Slow Brain module communicates with the router module to acquire multi-source situational information, perform task reasoning on this information, generate candidate solutions, and output them to the router module. The Slow Brain module is a high-level reasoning module, consisting of a task reasoning model, a knowledge base, a skill base, a tool library, and a memory module. It is primarily used to handle tasks requiring semantic understanding, knowledge retrieval, and complex situational awareness, such as task understanding, collaborative planning, and replanning under abnormal conditions.
[0024] It should be noted that, in this embodiment, the slow-brain module obtains multi-source situational information by receiving multi-source input data and standardizing it. This multi-source input data includes, but is not limited to, task instruction data, platform sensor data, and communication link status data. Task instruction data indicates the type and objective of the task to be performed by the underwater unmanned platform; platform sensor data reflects the underwater unmanned platform's own motion state, pose information, and equipment operating status; and communication link status data reflects the communication quality and link availability between the underwater unmanned platform and external systems.
[0025] Due to the diverse data sources in the underwater environment, data transmitted from different sensors and communication links may experience time delays, and may contain missing or conflicting information. To address these issues, the SlowBrain module can verify the received data based on the timestamps, confidence levels, and source priorities of each multi-source input data. Specifically, for multi-source input data with time delays, time-series alignment is performed based on timestamps; for missing multi-source input data, supplementation or labeling is performed based on source priority; and for conflicting multi-source input data, a comprehensive judgment and resolution are conducted based on confidence levels and source priorities. After the above verification process, the SlowBrain module integrates the multi-source input data into multi-source situational information in a unified format.
[0026] Next, the slow-brain module performs task reasoning on multi-source situational information to obtain candidate solutions. Specifically, in this embodiment, the slow-brain module can perform task reasoning on multi-source situational information based on a preset task reasoning model to obtain reasoning results, and then package the reasoning results to generate candidate solutions.
[0027] After obtaining a candidate solution, it is sent to the router module. The router module communicates with both the slow-brain module and the fast-brain module to receive the candidate solutions output by the slow-brain module, perform comprehensive judgment and security verification on the candidate solutions, and send the verified candidate solutions to the fast-brain module. The router module is an independent arbitration layer in this embodiment, located between the slow-brain module and the fast-brain module, and serves as a unified interface before the candidate solutions enter the execution link.
[0028] Specifically, after receiving a candidate solution, the router module first performs a comprehensive evaluation based on preset judgment conditions to obtain a judgment result. These preset judgment conditions include platform status conditions, communication quality conditions, and a credibility score for the candidate solution. The platform status condition determines whether the current state of the underwater unmanned platform allows the execution of the candidate solution; the communication quality condition determines whether the current communication link meets the communication requirements for the execution of the candidate solution; and the credibility score determines whether the credibility of the candidate solution itself meets the execution requirements.
[0029] It should be noted that the preset judgment conditions in this application embodiment are not limited to those listed above. Other judgment factors can be introduced according to actual task requirements and environmental changes, such as task stage conditions, target confidence conditions, external warning area conditions, and environmental risk conditions. Furthermore, the preset judgment conditions in this application embodiment are not static, but are flexibly adjusted according to changes in task stage, platform status, or feedback from the QuickBrain module, so as to make more accurate judgments on candidate solutions in different scenarios.
[0030] When the judgment result is satisfactory, the router module performs security verification on the candidate solution and obtains the verification result. Security verification includes rule verification, permission verification, security boundary verification, and resource constraint verification. Rule verification is used to verify whether the candidate solution conforms to the preset behavioral rules and task specifications; permission verification is used to verify whether the candidate solution has the permission to execute; security boundary verification is used to verify whether the candidate solution will trigger the platform's security boundary; resource constraint verification is used to verify whether the candidate solution exceeds the current available margin of energy, equipment, and other resources.
[0031] If the verification result is satisfactory, the router module will send the candidate solution to the QuickBrain module. If either the comprehensive judgment or the security verification fails, the router module will reject the candidate solution and notify the SlowBrain module to regenerate a candidate solution.
[0032] In this embodiment of the application, after the router module sends the verified candidate solution to the QuickBrain module, the QuickBrain module receives the verified candidate solution and generates a real-time control decision based on the verified candidate solution.
[0033] The QuickBrain module is the underlying real-time control module in this application embodiment. It consists of a behavior tree execution engine, a state machine, a small model reasoning, a local planning algorithm, a control algorithm, and safety rules. The QuickBrain module runs continuously online and is responsible for tasks that require rapid response, such as navigation control and local collision avoidance of underwater unmanned platforms. It does not involve complex semantic reasoning.
[0034] It should be noted that the small model inference built into the fast brain module can directly generate local real-time control decisions based on the current multi-source situation information when the slow brain module fails to generate candidate solutions or cannot respond in time. This allows the fast brain module to complete emergency control tasks independently without relying on the slow brain module in case of emergencies, such as scenarios like collision avoidance of sudden obstacles or recovery from safety boundary violations.
[0035] Specifically, after receiving a validated candidate solution, the QuickBrain module first parses the solution to obtain its structured intent. Then, based on a preset behavior tree template, the QuickBrain module translates the structured intent into an executable sequence of actions, which consists of a series of ordered behavior tree nodes. Finally, based on a preset control algorithm, the QuickBrain module maps each node in the action sequence, transforming the expected execution result of each node into specific real-time control decisions, such as low-level control commands like thrust and rudder deflection angle.
[0036] It should be noted that after generating real-time control decisions, the QuickBrain module also distributes these decisions to the underwater unmanned platform for execution. Simultaneously, the QuickBrain module continuously monitors the execution status of the underwater unmanned platform, obtains the execution results, and feeds these results back to the router module. Upon receiving these results, the router module, if the result indicates execution failure, notifies the SlowBrain module to regenerate candidate solutions.
[0037] Based on the above, this application provides an underwater unmanned platform decision-making system based on fast and slow brain routing arbitration. The system includes a slow brain module, a fast brain module, and a router module. The slow brain module acquires multi-source situational information, performs task reasoning on the multi-source situational information, generates candidate solutions, and outputs them to the router module. The router module receives candidate solutions, performs comprehensive judgment and security verification on the candidate solutions, and sends the verified candidate solutions to the fast brain module. The fast brain module receives the verified candidate solutions and generates real-time control decisions based on the verified candidate solutions.
[0038] As can be seen, this application, by setting up a router module between the slow-brain module and the fast-brain module, ensures that candidate solutions generated by the slow-brain module undergo comprehensive judgment and security verification before entering the fast-brain module. Only candidate solutions that pass the verification can be sent to the fast-brain module, avoiding the security risks caused by unexecutable solutions being directly sent to the underwater unmanned platform for execution. Thus, the security and feasibility of the candidate solutions themselves are judged before being sent to the underwater unmanned platform for execution. Therefore, this application constructs an effective verification and interception mechanism before candidate solutions are sent for execution.
[0039] Furthermore, Figure 2 A flowchart illustrating the switching of dominance between fast and slow brains is provided in this application embodiment, combined with... Figure 2 As shown, the underwater unmanned platform decision-making system defaults to a slow-brain-dominated mode. In this mode, the slow-brain module acquires multi-source situational information and performs task reasoning to generate candidate solutions. After verification by the router module, these solutions are sent to the fast-brain module. The fast-brain module generates real-time control decisions based on the verified candidate solutions and sends them to the underwater unmanned platform for execution. Simultaneously, the fast-brain module continuously feeds back the execution results to the router module.
[0040] After receiving the execution result, the router module adjusts the dominance of the fast and slow brains based on the result. Specifically, if the execution result indicates normal execution, the router module maintains the slow brain-dominated mode and continues the normal decision-making process. If the execution result indicates failure, the router module switches to the fast brain-dominated mode. In the fast brain-dominated mode, the fast brain module no longer relies on the candidate solutions of the slow brain module. Instead, based on its own acquired multi-source situational information and combined with built-in small model reasoning, it autonomously generates emergency real-time control decisions and sends them to the underwater unmanned platform for execution to ensure the safety of the underwater unmanned platform in abnormal situations. At the same time, the router module notifies the slow brain module to regenerate candidate solutions and temporarily stores the regenerated candidate solutions until the fast brain-dominated mode ends, at which point they are sent back to the fast brain module.
[0041] It should be noted that in the fast-brain-dominated mode, the slow-brain module is in a pending state. Specifically, the fast-brain-dominated mode can be applied to sudden abnormal scenarios, such as emergency ascent and emergency collision avoidance. Furthermore, the fast-brain-dominated mode can also be applied to routine, simple task scenarios, such as depth-keeping navigation and course-keeping navigation, which do not require high-level semantic reasoning. In these scenarios, the fast-brain module generates emergency real-time control decisions by acquiring multi-source situational information and using built-in small models to meet the real-time control decision-making needs of different scenarios. Once the execution state returns to normal, the router module switches the leadership back to the slow-brain-dominated mode, where the slow-brain module continues to generate candidate solutions and enters the normal decision-making chain.
[0042] Through the above process, the router module dynamically switches the leadership based on the real-time execution feedback of the QuickBrain module, enabling the underwater unmanned platform decision-making system to be autonomously taken over by the QuickBrain module in emergency control in abnormal situations based on multi-source situational information, thus ensuring the safe operation of the underwater unmanned platform.
[0043] Furthermore, Figure 3 A flowchart illustrating the candidate scheme degradation process under weak communication conditions is provided in this application embodiment, combined with... Figure 3 It is known that when an underwater unmanned platform is in a weak communication environment, the candidate solutions generated by the slow brain module based on the currently acquired multi-source situational information may include tasks that require coordination with other underwater unmanned platforms to complete, such as multi-platform collaborative tracking or distributed detection.
[0044] After receiving the candidate solution, the router module first checks the current communication quality. If the communication quality meets the requirements for cross-platform collaborative execution, the router module will normally send the candidate solution to the QuickBrain module for execution. If the communication quality does not meet the requirements for cross-platform collaborative execution, the candidate solution cannot be implemented under the current communication conditions. The router module will reject the candidate solution and notify the SlowBrain module to regenerate a single-platform execution solution that does not depend on cross-platform communication.
[0045] Next, the slow-brain module regenerates a single-platform candidate solution based on the rejection reason. After being re-verified by the router module, it is sent to the fast-brain module. Once communication is restored, the router module updates the communication quality status, and the underwater unmanned platform system can resume cross-platform collaborative planning. Through this process, the underwater unmanned platform system can proactively downgrade cross-platform collaborative solutions to single-platform execution solutions under weak communication conditions, avoiding mission risks caused by unexecutable solutions entering the execution link.
[0046] Furthermore, Figure 4 A flowchart illustrating a decision-making method for an underwater unmanned platform based on fast and slow brain routing arbitration, provided as an embodiment of this application. (Combined with...) Figure 4 As shown, this method is applied to the underwater unmanned platform decision-making system of the above embodiment.
[0047] S401: Through the slow brain module in the underwater unmanned platform decision system, it acquires multi-source situational information, performs task reasoning on the multi-source situational information, generates candidate solutions, and outputs them to the router module.
[0048] In this embodiment of the application, multi-source situational information is obtained through the slow brain module in the underwater unmanned platform decision-making system. The multi-source situational information is obtained by the slow brain module by receiving multi-source input data and performing standardized processing on the multi-source input data. The multi-source input data includes, but is not limited to, task instruction data, platform sensor data, and communication link status data.
[0049] It should be noted that, since multi-source input data comes from diverse sources, there may be issues such as time delays, missing content, and contradictions. Therefore, the embodiments of this application can standardize the multi-source input data in three aspects: time dimension, content dimension, and conflict handling, thereby obtaining multi-source situational data.
[0050] Specifically, for multi-source input data with time delays, time-series alignment is performed based on timestamps; for multi-source input data with missing information, supplementation or labeling is performed based on source priority; and for multi-source input data with conflicts, comprehensive judgment and resolution are performed based on confidence level and source priority.
[0051] Next, in this embodiment of the application, the multi-source situational information is processed by the task reasoning model preset in the slow brain module to obtain candidate solutions. After obtaining the candidate solutions, the candidate solutions are sent to the router module for subsequent judgment and verification.
[0052] S402: Receives candidate solutions through the router module in the underwater unmanned platform decision-making system, performs comprehensive judgment and security verification on the candidate solutions, and sends the verified candidate solutions to the QuickBrain module.
[0053] In this embodiment, after receiving the candidate solutions sent by the slow-brain module through the router module in the underwater unmanned platform decision-making system, the router module first performs a comprehensive judgment on the candidate solutions based on preset judgment conditions to obtain a judgment result. These preset judgment conditions include, but are not limited to, platform status conditions, communication quality conditions, and credibility scoring conditions for the candidate solutions.
[0054] It should be noted that after the candidate solution passes the comprehensive evaluation, the router module further performs security verification on the candidate solution to obtain the verification results. Security verification includes rule verification, permission verification, security boundary verification, and resource constraint verification. Rule verification verifies whether the candidate solution conforms to preset behavioral rules and task specifications; permission verification verifies whether the candidate solution has the necessary execution permissions; security boundary verification verifies whether the candidate solution will trigger the security boundary of the underwater unmanned platform; and resource constraint verification verifies whether the candidate solution exceeds the available margin of current energy, equipment, and other resources.
[0055] If either the comprehensive evaluation or the security verification fails, the router module rejects the candidate solution and notifies the slow-brain module to regenerate a candidate solution. If both the comprehensive evaluation and the security verification pass, the router module sends the candidate solution to the fast-brain module.
[0056] S403: Receives verified candidate solutions through the QuickBrain module in the underwater unmanned platform decision-making system, and generates real-time control decisions based on the verified candidate solutions.
[0057] In this embodiment, after receiving the verified candidate solution from the router module, the QuickBrain module in the underwater unmanned platform decision-making system parses the candidate solution to obtain the structured intent within it. Then, based on a preset behavior tree template, the QuickBrain module translates the structured intent into an action sequence, and further maps the action sequence using a preset control algorithm to generate a real-time control decision.
[0058] It should be noted that the real-time control decisions generated by the Fast Brain module are sent to the underwater unmanned platform for execution. During execution, the Fast Brain module continuously feeds back the execution results to the Router module, which then adjusts the control of the Fast and Slow Brain modules based on the results.
[0059] Furthermore, Figure 5 This application provides a structural block diagram of an underwater unmanned platform decision-making method based on fast and slow brain routing arbitration, combined with... Figure 5As shown in the diagram, the structure includes a data input layer, a situation processing layer, a decision-making layer, and an execution layer.
[0060] Specifically, the data input layer receives multi-source input data, including but not limited to task command data, platform sensor data, and communication link status data. The situation processing layer contains a standardization processing unit, which standardizes the multi-source input data received by the data input layer to form unified multi-source situation information, which is then input to the decision layer.
[0061] The decision-making layer comprises a slow-brain module, a router module, and a fast-brain module. The slow-brain module receives multi-source situational information, performs task reasoning based on this information, generates candidate solutions, and outputs them to the router module. The router module receives candidate solutions, performs comprehensive evaluation and security verification on them, and sends verified candidate solutions to the fast-brain module. The fast-brain module receives verified candidate solutions, generates real-time control decisions based on these decisions, and sends these decisions to the execution layer.
[0062] The execution layer is an underwater unmanned platform used to receive real-time control decisions from the fast-brain module and execute corresponding actions, including propulsion, servo, payload, and communication operations. It also generates execution results, including but not limited to execution records, abnormal events, and status feedback. Simultaneously, the execution results are fed back to the decision layer to trigger the slow-brain module to regenerate candidate solutions or the router module to update its judgment results.
[0063] It should be noted that there is bidirectional interaction between the slow-brain module and the router module, and between the router module and the fast-brain module. The slow-brain module outputs candidate solutions to the router module, the router module sends the verified candidate solutions to the fast-brain module, and the fast-brain module feeds back the execution results to the router module. The feedback layer feeds back the execution results from the execution layer to the decision layer, forming a complete closed-loop control link.
[0064] Furthermore, embodiments of this application also provide an electronic device, including: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, the one or more programs including instructions that, when executed by the processor, cause the processor to perform any of the implementation steps of the underwater unmanned platform decision-making method based on fast and slow brain routing arbitration described above.
[0065] Furthermore, embodiments of this application also provide a computer-readable storage medium for storing a computer program, which, when executed by a terminal device, implements any of the implementation steps of the underwater unmanned platform decision-making method based on fast and slow brain routing arbitration.
[0066] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that all or part of the steps in the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application. It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on describing the differences from other embodiments. The same or similar parts between the various embodiments can be referred to mutually.
[0067] The system disclosed in the embodiments is described in a relatively simple manner because it corresponds to the method disclosed in the embodiments. For relevant details, please refer to the method section.
[0068] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0069] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A decision-making system for an underwater unmanned platform based on fast and slow brain routing arbitration, characterized in that, The system includes a slow brain module, a fast brain module, and a router module; The slow brain module is used to acquire multi-source situational information, perform task reasoning on the multi-source situational information, generate candidate solutions, and output them to the router module. The router module is used to receive the candidate schemes, perform comprehensive judgment and security verification on the candidate schemes, and send the verified candidate schemes to the QuickBrain module. The fast brain module is used to receive the verified candidate solutions and generate real-time control decisions based on the verified candidate solutions.
2. The system according to claim 1, characterized in that, The fast brain module is also used to send the real-time control decision to the underwater unmanned platform for execution, obtain the execution result of the underwater unmanned platform, and feed the execution result back to the router module; The router module is also used to receive the execution result, and when the execution result indicates that the execution failed, to notify the slow brain module to regenerate the candidate solution.
3. The system according to claim 1, characterized in that, The step of comprehensively judging and verifying the candidate solutions, and sending the verified candidate solutions to the QuickBrain module, includes: Based on preset judgment conditions, the candidate solutions are comprehensively judged to obtain the judgment result; When the determination result is passed, the candidate solution is subjected to security verification to obtain the verification result. The security verification includes rule verification, permission verification, security boundary verification and resource constraint verification. When the verification result passes, the candidate solution is sent to the QuickBrain module.
4. The system according to claim 1, characterized in that, The process of generating real-time control decisions based on the verified candidate schemes includes: The candidate schemes that pass the verification are parsed to obtain the structured intent; Based on a preset behavior tree template, the structured intent is translated to obtain an action sequence; Based on a preset control algorithm, the action sequence is mapped to generate real-time control decisions.
5. The system according to claim 1, characterized in that, The step of performing task reasoning on the multi-source situational information to generate candidate solutions includes: Based on a preset task reasoning model, task reasoning is performed on the multi-source situational information to obtain the reasoning result; The reasoning results are packaged to generate the candidate solutions.
6. The system according to claim 1, characterized in that, The acquisition of multi-source situational information includes: Receive multi-source input data, including task instruction data, platform sensor data, and communication link status data; The multi-source input data is standardized to obtain the multi-source situational information.
7. The system according to claim 3, characterized in that, The preset judgment conditions include platform status conditions, communication quality conditions, and credibility scoring conditions. The platform status condition is used to determine whether the status of the underwater unmanned platform allows the execution of the candidate scheme; The communication quality conditions are used to determine whether the communication link meets the communication requirements required for the execution of the candidate scheme. The credibility scoring criteria are used to determine whether the credibility of the candidate solution meets the execution requirements.
8. A decision-making method for underwater unmanned platforms based on fast and slow brain routing arbitration, characterized in that, The underwater unmanned platform decision-making system according to any one of claims 1 to 7 includes: The slow brain module in the underwater unmanned platform decision-making system acquires multi-source situational information, performs task reasoning on the multi-source situational information, generates candidate solutions, and outputs them to the router module. The router module in the underwater unmanned platform decision-making system receives the candidate solutions, performs comprehensive judgment and security verification on the candidate solutions, and sends the verified candidate solutions to the QuickBrain module. The fast-brain module in the underwater unmanned platform decision-making system receives the verified candidate schemes and generates real-time control decisions based on the verified candidate schemes.
9. An electronic device, characterized in that, The device includes: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform the steps of the underwater unmanned platform decision-making method based on fast and slow brain routing arbitration as described in claim 8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program, which, when executed by a terminal device, implements the steps of the underwater unmanned platform decision-making method based on fast and slow brain routing arbitration as described in claim 8.
