Distributed fault detection and dynamic task scheduling system for unmanned ship cluster

Abstract

The invention provides a distributed fault detection and dynamic task scheduling system for an unmanned ship cluster, and the system achieves the efficient cooperation of the unmanned ship cluster in a complex marine environment through the cooperation of a sensor data collection module, a data processing and feature extraction module, a fault state classification module, an information sharing module, a weighted voting decision module, a task scheduling module and the like. The system firstly collects state information of each unmanned ship and adjacent ships, performs feature extraction and classification on data, and judges whether a fault exists or not. Fault information is shared to each boat in a cluster through a multi-path redundancy communication mechanism, and weighted decision is carried out by a voting mechanism, so that the accuracy of fault detection is improved. After a fault is confirmed, the system dynamically adjusts task allocation according to task priorities, key tasks are transferred to the unmanned surface vehicle in a good state, and continuity and stability of cluster tasks are guaranteed. Different from traditional centralized control, the distributed control method disclosed by the invention enhances the damage resistance and the self-adaptability of the system.

Description

Technical Field

[0001] The present invention relates to unmanned watercraft swarm technology, and in particular to a method and system for implementing distributed fault detection and dynamic task scheduling in a swarm, aiming to improve the communication reliability and task execution robustness of the unmanned watercraft swarm in complex environments. Background Art

[0002] In modern maritime warfare and civilian operations, the coordinated operation of swarms of unmanned aerial vehicles (UAVs) has become crucial. By integrating multiple unmanned platforms, UAV swarms can undertake complex and diverse missions, such as collaborative situational awareness, formation escort, search and rescue, resource exploration, and swarm confrontation. However, the complexity of the marine environment and limited communication conditions make traditional centralized control approaches susceptible to single points of failure and communication delays, which can seriously affect the efficiency and success rate of mission execution. Centralized control methods rely on a central node to coordinate and allocate tasks, which is responsible for unified command of all UAV operations. Under good communication conditions, the central node can efficiently coordinate swarm tasks. However, in harsh marine environments, communication links may be disrupted or delays may prevent timely transmission of commands, thereby reducing the system's response speed. Furthermore, if the central node fails, the coordinated operation of the entire swarm may be immediately interrupted, significantly increasing the risk of mission failure. To overcome these problems, distributed control methods have emerged as an effective solution. Distributed control methods allow each UAV to act as an independent intelligent agent with autonomous decision-making capabilities, share information with other UAVs, and collaborate on missions. This not only reduces reliance on central communication nodes and avoids mission interruptions due to single points of failure, but also significantly improves the robustness of the system, enabling the UAV swarm to complete its mission even if individual vessels fail. Furthermore, the distributed control architecture enhances the system's resilience and adaptability, making it more adaptable to harsh marine environments. Within this distributed control framework, UAV swarms can employ a variety of collaborative control strategies, such as behavioral and consistency methods, to achieve formation control and task allocation. These strategies not only maintain swarm network connectivity but also optimize communication power, reduce internal interference, and enhance the swarm's resilience. Furthermore, to adapt to the dynamic changes in the marine environment, UAV swarms also require cross-domain collaborative control capabilities, including key technologies such as collaborative communication, formation control, and mission planning. The combination of these capabilities enables UAV swarm systems to efficiently and stably perform missions in complex environments, offering broad application prospects. Summary of the Invention

[0003] (1) Technical problems solved

[0004] The technical problem solved by the present invention is how to realize distributed fault detection and dynamic task scheduling of unmanned boat clusters under communication delay and complex environmental conditions, thereby improving the robustness of the system and the efficiency of task execution.

[0005] (2) Technical solution

[0006] A distributed fault detection and dynamic task scheduling system for unmanned boat clusters includes the following modules: a sensor acquisition module, a data processing and feature extraction module, a machine learning classification module, a redundant communication and information sharing module, a voting decision module, and a task scheduling module.

[0007] Among them, the sensor acquisition module is used to collect the status information of neighboring unmanned boats in real time, providing basic data for fault detection and task decision-making; the data processing and feature extraction module processes the collected raw data, extracts features to generate behavioral feature vectors; the machine learning classification module determines whether the unmanned boat has a fault by analyzing the feature data and generates a status label; the redundant communication and information sharing module ensures the reliable sharing of fault information within the cluster through a multi-path communication mechanism; the voting decision module uses a weighted majority voting mechanism to confirm the fault status based on the judgment results of each unmanned boat; after confirming the faulty unmanned boat, the task scheduling module dynamically assigns its tasks to other normally working unmanned boats to ensure task continuity and system stability.

[0008] The sensor acquisition module is the most basic data input module in the system, providing basic information for subsequent data processing and feature extraction modules. This module is closely connected with the data processing module and ensures the accuracy of fault detection and task scheduling by continuously transmitting real-time data. The sensor acquisition module is responsible for collecting the status information of each unmanned boat and its neighboring boats in real time, including data such as position, speed, and heading. Since unmanned boat clusters often operate in marine environments with unstable communications, the sensor acquisition module can adapt to harsh environments and maintain data stability and continuity. The collected data can not only be used for fault detection, but also provide basic information for task collaboration and cluster behavior adjustment.

[0009] The data processing and feature extraction module receives raw data from the sensor acquisition module and transmits the extracted feature vectors to the machine learning classification module as input for classification analysis. This module filters, smoothes, and de-noises the raw sensor data, extracting key behavioral features (such as speed fluctuations and position changes). Unmanned vehicle swarms often operate in dynamic and unpredictable environments, so this module generates features that reflect environmental changes and behavioral anomalies, ensuring that subsequent classification modules can more accurately identify faults. Furthermore, the module's design supports low-latency processing, allowing feature extraction to keep pace with data updates and meet real-time detection requirements.

[0010] The machine learning classification module uses the feature vectors generated by the data processing and feature extraction module to analyze the status of neighboring unmanned boats and generate fault or normal status labels. The output of this module will be passed to the redundant communication and information sharing module to share the detection results with other unmanned boats. The machine learning classification module uses a machine learning algorithm to classify the unmanned boat feature data, identify and mark the fault status. Due to the distributed characteristics of the unmanned boat cluster, the machine learning classification module runs independently within a single boat, allowing each unmanned boat to perform local fault detection and reduce dependence on the central node. Through the adaptive characteristics of the algorithm, the module can adapt to the changing conditions in the environment, thereby effectively identifying faulty boats and providing reliable fault labels for the voting decision module.

[0011] The redundant communication and information sharing module receives the fault label information from the machine learning classification module and transmits this information to other unmanned boats through a multi-path communication mechanism for collective judgment in the voting decision module. The redundant communication and information sharing module ensures that each unmanned boat can share fault status information with adjacent boats, and uses multi-path redundant communication to improve the reliability and anti-interference capability of data transmission. Given that the unmanned boat cluster may be in a low-bandwidth or high-latency marine environment, the redundant communication mechanism transmits information through multiple paths to reduce data loss and communication delay problems. This module can also perform repeated detection and verification of shared data, thereby improving the system's data sharing quality and task collaboration capabilities in harsh environments.

[0012] The voting module receives fault information from the redundant communication and information sharing module and uses a weighted majority vote to determine the final fault status of each unmanned vehicle. The voting results are fed back to the task scheduling module as a basis for task adjustments. The voting module collectively determines the fault status using a weighted majority voting mechanism. The voting weight of each unmanned vehicle is adjusted in real time based on its communication quality, observation accuracy, and other factors, ensuring that high-quality information is prioritized in the decision-making process.

[0013] The task scheduling module dynamically allocates tasks based on the results of the voting decision module. Upon confirming a failure on a UAV, its tasks are promptly transferred to other normally functioning UAVs, ensuring the continuity and stability of the cluster's mission. The task scheduling module dynamically adjusts task allocation based on the health status of the UAVs and mission requirements. Specifically, upon detecting a failure, high-priority tasks are preferentially assigned to UAVs with better communication quality and observation performance. The task scheduling module supports cross-domain task transfer, ensuring efficient collaboration and sustained mission execution in the cluster even in the presence of failures.

[0014] (3) Beneficial effects

[0015] Compared with the existing technology, the present invention provides a distributed fault detection and dynamic task scheduling system for unmanned boat swarms, which has the following beneficial effects:

[0016] 1. Improve the robustness and fault tolerance of the system: Through distributed fault detection and voting decision-making mechanisms, each unmanned boat can independently detect the status of neighboring boats, avoiding the single point of failure problem of traditional centralized control. Even if a single boat or a few boats fail, the cluster can still operate stably and ensure mission continuity.

[0017] 2. Achieve efficient scheduling and dynamic adjustment of tasks: After confirming the faulty UAV, the system can adjust task allocation in real time based on the task priority and the communication status of the UAV, giving priority to assigning high-priority tasks to UAVs in good condition, ensuring that critical tasks are executed first and improving the overall completion rate of cluster tasks.

[0018] 3. Improved system adaptability and resilience: Through an intelligent voting weighting mechanism and dynamic task scheduling, each unmanned vehicle can make independent decisions based on its own observation data and that of neighboring vehicles, reducing reliance on central nodes. The system can adaptively adjust task allocation and coordination strategies in response to dynamic changes in the external environment, maintaining efficient collaboration within the swarm.

[0019] 4. Adaptability to Diverse Mission Requirements: The system architecture and modular design of this invention are highly flexible and applicable to a variety of complex mission scenarios, including formation control, resource exploration, and collaborative detection. The distributed modular design makes the system easy to expand and optimize, providing highly reliable support for diverse mission requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1 It is a structural schematic diagram of the present invention; DETAILED DESCRIPTION

[0021] To facilitate understanding of the present application, the present application will be described in more detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that all technical and scientific terms used in this specification have the same meaning as those commonly understood by those skilled in the art in the technical field of this application. The terms used in this specification and in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" used in this specification includes any and all combinations of one or more related listed items.

[0022] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other. In the actual application process, the following steps are specifically included:

[0023] Step 001. Each unmanned boat collects status information about itself and its neighbors through sensors. The required data includes key parameters such as position, speed, and heading. The collection equipment includes GPS locators, accelerometers, gyroscopes, ultrasonic sensors, or lidar.

[0024] Step 002: The collected data is further processed, including smoothing and denoising, to generate behavioral feature vectors. These feature data, such as position change rate, speed fluctuation, and heading change, can reflect the motion characteristics of the unmanned vehicle.

[0025] Step 003. The extracted behavioral features are input into a machine learning model to analyze the UAV's operating status. The classification algorithm used here can be logistic regression or a gradient boosted decision tree (GBDT). With the support of the pre-trained model, the system determines whether the UAV exhibits abnormal behavior and labels the analysis result as "faulty" or "normal."

[0026] Step 004: Share the fault status tag with other UAVs in the cluster via a multi-path redundant communication mechanism. This redundant communication mechanism transmits information over multiple transmission paths to reduce the risk of data loss and transmission delays.

[0027] Step 005. After receiving the fault information, each UAV collectively determines the fault status. This voting mechanism uses a weighted majority voting method, and the voting weight of each UAV is dynamically adjusted based on factors such as communication quality and observation accuracy.

[0028] Step 006. After the fault is confirmed, the tasks of the faulty UAV are dynamically adjusted. Based on the priority and current status of the tasks, the tasks are reallocated to other UAVs in the cluster to ensure that critical tasks are executed first.

[0029] In real-world scenarios, the UAV's behavioral characteristics are input into the model, which then outputs a label representing whether it is faulty or not, such as "normal" or "faulty." In specific cases, to improve recognition robustness, a multi-classifier model can be introduced, combining the results of different algorithms to form a comprehensive judgment. For example, if the system detects behavioral characteristics such as an abnormal rate of position change or unstable heading, it will label it as "faulty." Furthermore, state confidence information is generated to provide a reference value for decision-making, assisting other UAVs in the swarm in making effective state judgments.

[0030] In most communication scenarios, we cannot accurately and precisely transmit information. By accessing a redundant communication mechanism, the probability of communication interruptions can be greatly reduced. The specific implementation of the redundant communication mechanism includes: First, when the unmanned boat sends fault information, it selects multiple transmission paths to ensure that when one path is blocked, the information can still be transmitted on other paths; second, the path selection is based on real-time environmental monitoring data, such as signal strength and noise interference, to ensure the optimal path transmission efficiency. The mechanism further designs an information confirmation feedback process, that is, the receiving boat sends confirmation feedback after receiving the information, so that the sending boat can determine whether the transmission is successful and decide whether to send it again. This multi-path redundant transmission and confirmation mechanism jointly ensures that fault information is quickly and reliably transmitted to other members of the cluster in complex environments.

[0031] In the weighted voting described above, weights are adjusted based on multiple factors, such as communication quality, observation accuracy, and historical judgment accuracy. For example, when a UAV's communication quality is poor or the sensor observation environment is poor, its voting weight will automatically decrease to prevent its misjudgment from interfering with collective decision-making. The specific steps of the weighted voting mechanism include: 1. After the fault information is shared with the cluster, each UAV forms a judgment of "fault" or "normal" based on its own observation conditions and judgment capabilities. 2. The cluster system weights and accumulates the judgment results of each UAV, and judgments with higher weights have a greater impact on the final result. 3. The system sets a fault confirmation threshold for the weighted voting results, and if the threshold is exceeded, it is considered a collective confirmation of the fault state. For example, when the majority of UAVs vote "fault" with a higher weight, the system will mark it as a fault confirmation and then enter the next step of the task adjustment process.

[0032] The present invention implements dynamic task scheduling for the specifically designed dynamic task adjustment method described above through the following steps: 1. Task priority analysis: First, the tasks of the faulty unmanned vehicle are prioritized. Critical tasks (such as data acquisition and target monitoring) are marked as having a higher priority. Non-critical tasks are marked as having a lower priority based on their urgency. 2. Reorganization of normal unmanned vehicle tasks: Based on the current status, task load, and location of the normal unmanned vehicles in the cluster, an appropriate unmanned vehicle is selected to take over the high-priority tasks of the faulty unmanned vehicle. The selection strategy here includes dynamic sorting based on factors such as the unmanned vehicle's remaining energy, distance from the task point, and current task load to ensure that task allocation conforms to the actual environment and resource conditions. 3. Real-time dynamic adjustment: During task execution, the system continuously monitors the status of the unmanned vehicle and the completion of the task, and dynamically adjusts task allocation. If a new fault occurs in the assigned unmanned vehicle or the environment changes significantly, the system will reanalyze the resource status in the cluster and reallocate tasks to ensure that the cluster can flexibly adapt to emergencies.

Claims

1. A distributed fault detection and dynamic task scheduling system for unmanned boat clusters, comprising a sensor acquisition module, a data processing and feature extraction module, a machine learning classification module, a redundant communication and information sharing module, a voting decision module and a task scheduling module, characterized in that: S1: The sensor acquisition module is used to collect the status information of each unmanned boat and its neighboring boats in real time, and the status information includes parameters such as position, speed and heading; S2: the data processing and feature extraction module performs data preprocessing on the state information, extracts behavior features, and generates feature vectors; S3: The machine learning classification module classifies the state of the unmanned boat using a pre-trained machine learning algorithm based on the feature vector, and outputs a "fault" or "normal" state label; S4: The redundant communication and information sharing module shares fault information within the cluster through a multi-path redundant communication method to ensure that the fault information can be effectively transmitted to other unmanned boats when the communication conditions are poor; S5: The voting decision module performs weighted majority voting based on the fault information, and the voting weight of each unmanned boat is dynamically adjusted according to factors such as communication quality and observation accuracy to confirm the final fault status; S6: After confirming the fault, the task scheduling module dynamically adjusts the task of the faulty unmanned boat and assigns the task to the unmanned boat in normal state to ensure the continuous execution of the task.

2. The system according to claim 1, characterized in that The sensor acquisition module includes a GPS locator, an accelerometer, a gyroscope, an ultrasonic sensor and / or a laser radar, which is used to accurately collect real-time status information of the unmanned boat.

3. The system according to claim 1, characterized in that The data processing and feature extraction module includes data smoothing, denoising and feature calculation units, which are used to filter, reduce noise and extract features from the collected state information.

4. The system according to claim 1, characterized in that The machine learning classification module uses logistic regression or gradient boosted decision tree (GBDT) algorithm to improve the accuracy of fault detection.

5. The system according to claim 1, characterized in that The redundant communication and information sharing module realizes reliable data transmission through multi-hop transmission and data verification mechanism, and reduces information loss caused by communication delay or interference.

6. The system according to claim 1, characterized in that The voting decision module dynamically adjusts the voting weight based on the communication quality and data integrity of the unmanned boats, and gives priority to adopting information from unmanned boats with better observation conditions.

7. The system according to claim 1, characterized in that The task scheduling module prioritizes task recipients according to task priorities and the real-time status of the unmanned boat to ensure efficient execution of key tasks.

Images