An open architecture system for an underwater acoustic signal processor

Abstract

The application discloses an open architecture system for a water acoustic signal processor, and belongs to the technical field of water acoustic signal processors.The system comprises a hardware resource pool, a standardized algorithm container group, a visual processing link construction module, a containerized deployment and scheduling platform, and a data processing and communication plane.The system pools special hardware resources, encapsulates water acoustic processing algorithms into standardized containers, allows users to arrange processing procedures in a graphical manner, dynamically deploys algorithm containers onto matched hardware for execution according to real-time resource conditions by means of an intelligent scheduling platform, and simultaneously provides an efficient and reliable data communication channel.The application solves the problems of poor expansibility, low resource utilization and the like of a traditional system, realizes elastic sharing of hardware resources, agile deployment of algorithms and flexible construction of procedures, and significantly improves the openness, flexibility and overall efficiency of a water acoustic signal processing system.

Description

Technical Field

[0001] This invention belongs to the field of underwater acoustic signal processing technology, and specifically relates to an open architecture system for underwater acoustic signal processing. Background Technology

[0002] Underwater acoustic signal processing technology is one of the core technologies in underwater exploration, communication, navigation, and marine environmental monitoring, and is widely used in many important industries such as national defense, marine scientific research, resource exploration, and underwater engineering. Traditional underwater acoustic signal processing systems typically adopt a closed, tightly coupled hardware and software design pattern, that is, for specific processing tasks or algorithms, a dedicated signal processing board or computing module is customized, and the algorithm is embedded into hardware logic or deeply bound to hardware embedded software.

[0003] As underwater acoustic applications become increasingly complex and diverse, with growing demands for multi-task parallel processing, rapid integration of novel adaptive algorithms, and cross-platform (e.g., collaborative processing across multiple ships, underwater vehicles, and buoy nodes), the existing technical architecture reveals the following major defects and shortcomings:

[0004] The system is closed, making expansion and upgrades difficult: Traditional dedicated processor architectures are rigid, hardware resources are fixed, and software and algorithms are highly dependent on specific hardware. When new processing algorithms need to be introduced or new task requirements need to be met, hardware modification or redesign is often required, resulting in long system upgrade cycles and high costs, making it difficult to adapt to the rapid iterative development of underwater acoustic technology.

[0005] Low resource utilization and lack of flexible scheduling: System hardware resources (such as different types of dedicated computing units) are usually statically allocated to fixed processing tasks and cannot be dynamically allocated according to changes in real-time task load. This results in some hardware resources being idle while others may be overloaded when the workload is uneven, leading to low overall resource utilization and an inability to achieve pooling and efficient sharing of computing resources.

[0006] Algorithm integration and deployment processes are cumbersome: The development and integration of new algorithms often require in-depth knowledge of the underlying system and are tightly coupled with hardware drivers and operating systems. The deployment process involves complex cross-compilation, environment configuration, and manual installation, which is not only inefficient but also prone to errors, making it difficult to achieve rapid algorithm verification, standardized encapsulation, and large-scale reuse.

[0007] The construction of the processing pipeline is not intuitive and lacks visualization and verification methods: complex underwater acoustic signal processing flows typically consist of multiple algorithm modules connected in series or parallel. In existing technologies, building and modifying such processing pipelines mostly rely on code writing or configuration file modification. The process is not intuitive enough, and it is difficult to effectively verify and simulate the correctness of the pipeline logic, the compatibility of data interfaces, and performance before deployment.

[0008] Cross-node collaboration and data communication efficiency bottlenecks: In distributed processing scenarios involving multiple computing nodes, existing systems lack an efficient and unified data communication plane. Data exchange between nodes often relies on customized communication protocols, making it difficult to guarantee the real-time and reliable transmission of data streams (especially those requiring high throughput or low latency) between different algorithm modules, thus becoming a bottleneck for improving system performance.

[0009] Poor adaptability to dynamic heterogeneous environments: The existing architecture has difficulty in seamlessly integrating newly added heterogeneous computing nodes with different hardware capabilities (such as computing units from different ships), and lacks the ability to perceive and adapt to dynamic network topology and time-varying link quality (such as underwater acoustic channels), which limits the flexibility and robustness of the system in mobile, multi-platform networking scenarios.

[0010] Therefore, there is an urgent need in this field for a new underwater acoustic signal processing system architecture to solve the aforementioned problems of being closed, rigid, inefficient, difficult to expand and maintain. Summary of the Invention

[0011] In view of this, the purpose of this invention is to provide an open architecture system for underwater acoustic signal processors, which realizes elastic pooling of hardware resources, standardized encapsulation and agile deployment of processing algorithms, visual construction and verification of processing links, and intelligent scheduling and unified management of system resources, thereby comprehensively improving the flexibility, scalability, resource utilization and overall performance of underwater acoustic signal processing systems.

[0012] To achieve the above objectives, the present invention provides the following technical solution:

[0013] An open architecture system for underwater acoustic signal processing includes:

[0014] The hardware resource pool includes multiple computing nodes containing dedicated hardware units for underwater acoustic signal processing. These computing nodes are connected through an interconnection network to form a set of computing resources for dynamic resource allocation.

[0015] The standardized algorithm container group includes multiple independently packaged containerized underwater acoustic signal processing algorithm units. Each algorithm unit packages the preset algorithm code, dependency libraries and configuration files into a uniform format image, and declares the type and number of dedicated hardware units required for its operation.

[0016] The visualization processing link building module is used by users to connect and configure parameters of multiple algorithm units in a standardized algorithm container group according to preset processing logic, define a complete underwater acoustic signal processing flow, and generate a flow description file that can be executed by a containerized deployment and scheduling platform.

[0017] The containerized deployment and scheduling platform is used to obtain process description files from the visual processing link building module, and instantiate and schedule algorithm containers to the matching dedicated hardware units for execution based on the needs of real-time processing tasks and the availability of hardware resource pools.

[0018] The data processing and communication plane establishes data flow channels between various algorithm container instances to transmit underwater acoustic data, including raw acoustic data, intermediate processing results, and final situational information.

[0019] As a further preferred embodiment of the present invention, the containerized deployment and scheduling platform includes:

[0020] The resource status monitoring module continuously collects load, utilization and health status data of each dedicated hardware unit in the hardware resource pool to form a global resource view;

[0021] The process parsing and scheduling module receives the process description file generated by the visualization processing link construction module, parses out the algorithm container dependencies, resource requirements and data flow contained therein, and, based on the global resource view, uses a comprehensive evaluation algorithm to allocate corresponding computing nodes and dedicated hardware units to each algorithm container instance.

[0022] The container lifecycle management module executes the instantiation, startup, shutdown, and resource reclamation operations of algorithm containers on the target computing node specified by the process parsing and scheduling module, and provides runtime status monitoring and fault recovery functions for running container instances.

[0023] As a further preferred embodiment of the present invention, the process parsing and scheduling module performs resource matching and scheduling in the following manner:

[0024] When parsing the process description file, identify the dedicated hardware unit requirements declared for each algorithm container;

[0025] In the global resource view, a set of candidate computing nodes that currently meet the dedicated hardware requirements and have sufficient available resources is selected;

[0026] If there are multiple candidate nodes, the optimal node will be selected for deployment based on a comprehensive evaluation of each node's real-time load, data communication overhead with upstream and downstream containers, and historical task execution efficiency.

[0027] If no node currently meets the mandatory hardware requirements of all containers, a resource shortage alarm is returned to the visualization processing link building module, or an alternative algorithm is executed on the general computing unit according to a preset strategy.

[0028] As a further preferred embodiment of the present invention, the data processing and communication plane performs the following operations:

[0029] Based on shared memory and remote direct data access (RDMA) technology, cross-process or cross-node data channels are established between algorithm container instances located on the same node or different nodes.

[0030] The data distribution model adopts a publish-subscribe approach. Algorithm container instances act as data producers, publishing results to specified topics. Algorithm container instances that need this data as input subscribe to the corresponding topics, and the communication plane completes the real-time forwarding and routing of data.

[0031] Configure differentiated transmission quality of service strategies for different types of underwater acoustic data streams, including ensuring that the throughput of the broadband stream data after beamforming is greater than a preset throughput threshold, and ensuring that the latency of the target tracking command data is less than a preset latency threshold.

[0032] As a further preferred embodiment of the present invention, the visualization processing link construction module includes:

[0033] An interactive orchestration canvas graphically displays selectable algorithm container units, which can be used to build processing chains from raw data input to final output through drag-and-drop and connection operations;

[0034] The link logic verification unit automatically checks whether the input and output data formats, sampling rates, and number of channels between adjacent algorithm container units in the link are compatible during the orchestration process, and provides modification prompts.

[0035] Run the simulation unit to inject historical or simulation data, simulate the execution process of the constructed processing link in an offline environment, and preview the data processing results and performance indicators of preset key links.

[0036] As a further preferred embodiment of the invention, it also includes a unified maintenance and management plane, which includes:

[0037] The algorithm image repository service is used for centralized storage, version management, and distribution of all algorithm images of the standardized algorithm container group. It supports users to submit new version images and records each update log.

[0038] The system configuration and monitoring center is used to centrally configure the network parameters of the hardware resource pool, the strategies of the data processing and communication plane, and monitor the overall operating status, performance indicators and alarm information of the entire system and each algorithm container.

[0039] As a further preferred embodiment of the present invention, it also includes a dynamic node management and authentication module, used to detect and identify newly added ship computing nodes through underwater acoustic signals, verify their identity and capabilities, and complete their registration or deregistration in the hardware resource pool; the dynamic node management and authentication module specifically performs the following operations:

[0040] Listen to the preset underwater acoustic signaling frequency band and capture the underwater acoustic authentication beacon signal broadcast by the new node, which contains its node ID and hardware configuration;

[0041] The captured underwater acoustic authentication beacon signal is decoded and its features are analyzed to verify whether the node ID exists in the preset list of trusted nodes;

[0042] If authentication is successful, the network access test of the new node is completed, and its hardware resource configuration information is updated in the global resource view of the containerized deployment and scheduling platform so that it can be used for task scheduling.

[0043] As a further preferred embodiment of the present invention, the containerized deployment and scheduling platform also includes:

[0044] The mobility and link quality awareness module is used to continuously evaluate the relative position changes between the computing nodes of each ship, the quality of the underwater acoustic channel, and the status of the radio frequency link, to obtain the link quality index and predict its changing trend.

[0045] When performing resource matching, the process parsing and scheduling module prioritizes computing nodes with a predicted link quality index higher than a preset value and a communication latency lower than a preset value for container deployment.

[0046] As a further preferred embodiment of the invention, a multi-layered security isolation and access control plane is also included, the plane comprising:

[0047] The containerized security sandbox module is used to configure resource access control lists and system call whitelists for each algorithm container instance, and to perform real-time monitoring and auditing during the runtime of the algorithm container instance.

[0048] The dynamic security classification module is used to dynamically label the security classification of data packets based on data source attributes or processing stages. The security classification-based access control gateway is located between the data channels of container instances and controls the forwarding and blocking of data flows based on the security classification labels and the permission level of the target container.

[0049] The beneficial effects of this invention are as follows:

[0050] This invention provides an open architecture system for underwater acoustic signal processors. By introducing key technologies such as hardware resource pooling, algorithm containerization, visual orchestration, intelligent scheduling, and unified data communication, it effectively overcomes the shortcomings of existing systems, specifically in the following aspects:

[0051] The system achieves openness and scalability, improving the efficiency of algorithm integration and deployment: By constructing a hardware resource pool containing heterogeneous dedicated hardware units, physical computing resources are abstracted into dynamically allocable pooled resources, breaking the traditional binding relationship between dedicated hardware and fixed tasks and providing a foundation for the access of new hardware. By defining standardized algorithm container groups, algorithms and their runtime environments are packaged into images of a unified format, achieving standardized algorithm encapsulation. This significantly simplifies the integration process of new algorithms, allowing algorithm developers to focus on underlying hardware details without needing to concern themselves with them, achieving "encapsulate once, run everywhere," and greatly improving the agility and reusability of algorithm deployment. The unified maintenance and management plane provides an algorithm image repository service, further supporting centralized storage, version management, and distribution of algorithms, facilitating continuous algorithm iteration and team collaboration.

[0052] Improved hardware resource utilization and system flexibility: The containerized deployment and scheduling platform, along with its included resource status monitoring module and process parsing and scheduling module, can dynamically and intelligently schedule algorithm container instances to the most suitable dedicated hardware units for execution based on real-time task requirements and a global resource view. This enables on-demand allocation and elastic scaling of computing resources, solving the problem of low utilization caused by static resource allocation, and allowing the system to efficiently handle varying task loads.

[0053] It provides an intuitive and reliable means of building and verifying processing links: a visual processing link building module and its interactive orchestration canvas allow users to build processing flows through graphical drag-and-drop and connection methods, greatly reducing the technical threshold for process construction and making the processing logic clear at a glance. The module's built-in link logic verification unit and runtime simulation unit can automatically check data interface compatibility before deployment and perform offline simulations using historical or simulation data to preview the results and performance of key links. This effectively avoids deployment failures due to design errors, shortens the development and debugging cycle, and improves the reliability and efficiency of system construction.

[0054] It ensures efficient and reliable cross-node data communication: The data processing and communication plane establishes efficient data channels based on technologies such as shared memory and RDMA, and adopts a publish / subscribe model for data distribution, providing a unified, high-performance data exchange infrastructure for distributed processing algorithm container instances. By configuring differentiated quality of service policies for different types of data streams (such as high-throughput broadband streams and low-latency instruction streams), it can meet the diverse real-time transmission needs in underwater acoustic signal processing, solving the bottlenecks of low efficiency and insufficient guarantee capabilities of traditional customized communication protocols.

[0055] Enhanced adaptability and collaborative processing capabilities in dynamic heterogeneous environments: The dynamic node management and authentication module can automatically detect, identify, and authenticate newly added heterogeneous computing nodes (such as other ship nodes) through underwater acoustic signals, and securely incorporate them into the hardware resource pool, achieving dynamic system expansion and plug-and-play functionality. The mobility and link quality awareness module can continuously evaluate the quality of communication links between nodes. When performing resource matching, the process parsing and scheduling module can prioritize the deployment of nodes with high link quality and low communication latency, thereby optimizing the overall performance and reliability of task execution and significantly improving the system's adaptability and robustness in dynamic and complex environments such as mobile and multi-platform networking.

[0056] The system's security and maintainability are enhanced: A multi-layered security isolation and access control plane, through containerized security sandboxes, dynamic security level labeling, and security level-based access control gateways, achieves strict isolation and controlled access between algorithms and data flows, effectively improving the overall system security. The unified maintenance and management plane's system configuration and monitoring center provides centralized configuration and panoramic monitoring capabilities for the entire system's hardware, network, and algorithm container runtime status, simplifying system operation and maintenance, facilitating rapid fault location and troubleshooting, and reducing maintenance costs.

[0057] Other advantages, objectives, and features of the invention will be set forth in the following description and will be apparent to those skilled in the art in some respects, or may be learned by practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0058] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration:

[0059] Figure 1 This is a schematic diagram of the overall architecture of the present invention;

[0060] Figure 2 This is a schematic diagram of the containerized deployment and scheduling platform of the present invention;

[0061] Figure 3 This is a schematic diagram of the visualization processing link construction module of the present invention. Detailed Implementation

[0062] like Figures 1-3 As shown, the core of this invention lies in constructing an open, scalable, and intelligently scheduled underwater acoustic signal processing system architecture. This system achieves a transformation from a fixed, closed traditional model to a flexible, open new model by pooling hardware resources, containerizing algorithms, and introducing a visual orchestration and intelligent scheduling platform. The following details the various components of the system and their collaborative working methods.

[0063] I. System Overall Framework and Workflow

[0064] The open architecture system of this invention mainly comprises the following five core components: a hardware resource pool, a standardized algorithm container group, a visual processing link construction module, a containerized deployment and scheduling platform, and a data processing and communication plane. Furthermore, the system preferably also includes a unified maintenance and management plane, a dynamic node management and authentication module, a mobility and link quality awareness module, and a multi-layered security isolation and access control plane to enhance the system's maintainability, dynamic adaptability, and security.

[0065] Its typical workflow is as follows:

[0066] Resource preparation and algorithm packaging: Administrators connect heterogeneous computing nodes (such as servers equipped with different models of digital signal processors (DSPs) and field-programmable gate arrays (FPGAs)) to the interconnection network to form a hardware resource pool. Algorithm developers package mature underwater acoustic signal processing algorithms such as beamforming, target detection, and feature extraction, along with their dependent library files and runtime configurations, into a unified container image (such as a Docker image), and label the required hardware type (such as "FPGA-A type") and quantity, storing them in a standardized algorithm container group.

[0067] Process Design and Validation: Task planners use the graphical interface of the visualization processing link construction module to drag and drop the required algorithm unit icons, such as "Data Preprocessing," "Beamforming," and "Target Tracking," from the algorithm container group onto the canvas, and define the data flow and configure parameters (such as beam scanning angle and detection threshold) by connecting them. This module automatically verifies the compatibility of data interfaces in the link. Subsequently, simulations can be run using historical data to preview the output results of key nodes. After confirming the correctness of the process, a structured process description file (such as YAML format) is generated.

[0068] Intelligent Scheduling and Deployment: The containerized deployment and scheduling platform receives the process description file. Its resource status monitoring module monitors the utilization rate of various hardware on each node in real time. The process parsing and scheduling module parses the file, identifies the hardware requirements of each algorithm container, and, combined with the global resource view and inter-node communication overhead, determines the optimal deployment node for each container (e.g., scheduling computationally intensive beamforming containers to idle FPGA nodes, and scheduling frequently interacting front-end and back-end containers to the same or adjacent nodes). The container lifecycle management module then pulls the corresponding algorithm image, instantiates it, and starts the container on the target node.

[0069] Data Stream Execution and Communication: Processing Task Initiation. Raw acoustic data is injected into the system through the data acquisition interface. The data processing and communication plane transmits data from the data source container to the downstream processing container in real time according to predefined publish / subscribe relationships. It utilizes shared memory (within the same node) and RDMA technology (across nodes) to ensure high throughput and low latency. For example, the broadband data stream after beamforming is marked as a high-throughput strategy to ensure its transmission bandwidth; while the command data generated by target tracking is marked as a low-latency strategy and transmitted with priority.

[0070] Dynamic Management and Security Control: In naval formation collaboration scenarios, new ship nodes can broadcast authentication beacons via underwater acoustic channels. After the dynamic node management and authentication module verifies their identity, their resources are added to the pool. The mobility and link quality awareness module continuously evaluates the link quality between nodes within the formation, allowing the scheduling platform to dynamically migrate containers to avoid links with degraded communication performance. Simultaneously, a security isolation and access control plane ensures that each algorithm container operates in a controlled "sandbox" and strictly controls data access between containers based on data security level labels.

[0071] II. Specific Execution of Each Module

[0072] 1. Core System Components

[0073] Hardware resource pool: Consists of multiple servers (computing nodes) interconnected via high-speed Ethernet or InfiniBand networks. Each server, in addition to a general-purpose CPU, is equipped with at least one dedicated hardware unit for underwater acoustic signal processing, such as a GPU for Fast Fourier Transform, an FPGA for real-time filtering, or an ASIC for accelerating specific algorithms. The resources of all nodes are virtualized into a unified resource pool. This breaks the fixed binding between hardware and tasks, enabling centralized management and elastic provisioning of computing resources, and laying the physical foundation for efficient resource reuse and dynamic scheduling.

[0074] Standardized Algorithm Container Groups: Utilizing containerization technologies (such as Docker), independent images are created for each algorithm. Each image contains the algorithm's executable file, dependent mathematical libraries (such as FFTW and CUDA libraries), system libraries, and a declaration file. The declaration file explicitly specifies the type and number of hardware accelerators required to run the image, as well as the computing resource quotas (such as the number of CPU cores and memory size). This achieves standardized and lightweight encapsulation of algorithms, resolving the strong coupling between algorithms and the underlying runtime environment (operating system, drivers). It allows algorithms to be quickly distributed, deployed, and versioned like software containers, greatly improving algorithm portability and reusability.

[0075] The Visual Processing Chain Building Module provides a graphical user interface (GUI). The left side displays the algorithm repository, showcasing all available containerized algorithm units and their brief descriptions. The main area is an orchestration canvas where users can build directed acyclic graphs (DAGs) by dragging and dropping lines. The right side is a properties panel for configuring the parameters of the selected algorithm unit. This transforms complex signal processing flow design from coding to intuitive graphical operation, significantly lowering the professional barrier and improving the efficiency and accuracy of flow design, enabling sonar operators without software development backgrounds to participate in flow customization.

[0076] Containerized Deployment and Scheduling Platform: This is a core backend service responsible for receiving task blueprints generated by the graphical interface and transforming them into actual, distributed containerized task execution. It needs to interact with the underlying container runtime (such as Docker Engine and containerd) and resource pools. It automates the process from task description to execution. Through dynamic scheduling, it optimizes global resource utilization, improving system throughput and task response speed.

[0077] Data Processing and Communication Plane: This is the data bus designed for high-frequency, high-capacity, real-time data streams in underwater acoustic signal processing. It is generated based on open-source message middleware (such as Apache Kafka, Pulsar) or high-performance communication libraries (such as Open MPI, UCX). It provides a unified, efficient, and reliable data exchange service, shielding the complexity of the underlying network and node locations. Through Quality of Service (QoS) policies, it ensures differentiated transmission guarantees for data streams with different service characteristics (high throughput / low latency), solving the data communication bottleneck in distributed processing.

[0078] 2. Intelligent resource matching of the scheduling platform

[0079] Resource Status Monitoring Module: A lightweight agent deployed on each computing node, which periodically collects metric data on the node's CPU, memory, and dedicated hardware (such as FPGA utilization and GPU memory usage), as well as the node's network bandwidth and health status (temperature, error count), and reports them to the central server of the scheduling platform to form a global resource view.

[0080] Process parsing and scheduling module: The implementation steps of this module are as follows:

[0081] Requirements analysis: Read the process description file and construct an internal task graph model, where nodes represent algorithm containers and edges represent data dependencies. Extract the specific hardware constraints (e.g., "must run on FPGA model X") and resource requests for each container.

[0082] Initial screening: Based on the global resource view, quickly filter out a list of candidate nodes that currently possess the required type and have sufficient remaining resources. For example, for a container requiring "FPGA-A type x1", only nodes with that type of FPGA installed and currently have at least one free unit will be included in the candidate list.

[0083] Comprehensive Evaluation and Optimization: If there is more than one candidate node, a scoring algorithm is activated. Scoring factors include: ① the node's current load (lower load is better); ② the estimated communication overhead with data producers or consumers (network latency, bandwidth; nodes with lower communication overhead are preferred, such as those on the same node or rack); ③ the node's historical success rate or efficiency in running similar tasks. A weighted scoring method is used to select the node with the highest total score.

[0084] Exception handling: If no node can meet the mandatory hardware requirements of a container, an explicit resource shortage warning is immediately returned to the user interface, with possible suggestions (such as waiting for resources to be released). Alternatively, the system can be configured with a degradation strategy, for example, after obtaining user confirmation, attempting to run a simplified version of the algorithm on a general-purpose CPU.

[0085] It achieves a leap from static allocation to dynamic intelligent matching. It not only meets the hard resource requirements, but also reduces communication latency and resource contention during task execution through multi-factor optimization, improves overall task execution efficiency and system throughput, and provides an elegant mechanism for handling insufficient resources.

[0086] The container lifecycle management module receives scheduling instructions (container image address, target node, resource configuration). It executes operations such as pulling images, creating containers, and starting them by calling the container runtime API (such as the Docker Daemon interface) on the target node. During runtime, it monitors the container status and attempts to restart or report failures.

[0087] 3. High-performance data communication plane

[0088] Channel Establishment: When the scheduling platform decides to deploy two data-dependent containers, the communication plane selects the optimal communication method based on their location information. If they are on the same node, a shared memory region is allocated to them; if they are on different nodes, a reliable, kernel-bypassed direct memory access channel is established between the two nodes through the RDMA driver.

[0089] Data distribution: The system maintains a topic registry. Upon startup, algorithm containers register their published (output) and subscribed (input) topics with the communication plane. For example, a beamforming container publishes the topic "beamformed-data," and an object detection container subscribes to this topic. The communication plane automatically routes data from publishers to all subscribers based on these subscription relationships.

[0090] QoS policy implementation: The communication plane configures transmission policies for different types of topics. For high-throughput topics such as "beamformed-data", large buffers and batch acknowledgment mechanisms are configured to maximize throughput. For low-latency topics such as "tracking-command", high-priority queues and immediate transmission of small packets are configured to ensure that end-to-end latency is below a set threshold (e.g., 10 milliseconds).

[0091] Shared memory and RDMA technologies significantly reduce data copying and protocol processing overhead, providing communication performance close to the hardware limits. The publish / subscribe model decouples data processing components, improving the system's loose coupling and scalability. Differentiated QoS policies ensure that the system can simultaneously meet the demands of large-volume data transmission and real-time control command delivery in underwater acoustic processing.

[0092] 4. Visualization of Building and Pre-validation

[0093] Interactive canvas arrangement: Implemented based on web technologies (such as HTML5 Canvas) or desktop frameworks (such as Qt). Supports graphical editing functions such as zooming, dragging, selection, and alignment.

[0094] Link Logic Verification Unit: This unit is automatically triggered each time a user adds or modifies a connection. It reads the metadata (defined during packaging) of the two connected algorithm container images and compares the output data format (such as floating-point arrays, complex matrices) with the input data format, sampling rate, and number of channels to see if they match. If they do not match, an error marker is highlighted on the connection line and a prompt box pops up.

[0095] Running the simulation unit: Users can click the "Simulation Run" button to upload or select a segment of historical measured data or simulated acoustic signal for the starting node of the processing link. Without actually scheduling the container, the system will call the standard implementation (or lightweight simulation model) of each algorithm unit in the background, perform rapid calculations according to the link sequence, and display intermediate or final result waveforms, spectrum diagrams, or performance statistics (such as processing time) at preset observation points on the canvas (such as chart components).

[0096] The design approach significantly improves user experience and design efficiency. Automated interface validation prevents runtime failures caused by data format mismatches from the outset. Offline simulation allows users to fully verify the logical correctness and estimated performance of the processing chain before investing real resources, reducing trial-and-error costs and project risks.

[0097] 5. Unified operation and maintenance management

[0098] Algorithm Image Repository Service: Deploy a private container image repository (such as Harbor). Algorithm developers push new versions of algorithm images to the repository via command-line tools. The repository manages image access permissions, version tags (such as "beamformer:v1.2"), and records metadata for each push (pusher, time, change summary).

[0099] System Configuration and Monitoring Center: Provides a centralized web management portal. Administrators can centrally configure the network addresses, VLAN information, and QoS policy templates for the communication plane for all compute nodes. The monitoring center displays a global resource utilization heatmap, real-time CPU / memory consumption and data traffic for each algorithm container, as well as a list of alarm events generated by the system through dashboards.

[0100] It achieves centralized asset management and global visibility of status. The mirror repository ensures algorithm version consistency and traceability. The configuration center simplifies the deployment and change management of large-scale systems. Comprehensive monitoring enables operations personnel to quickly grasp the system's health status, promptly locate performance bottlenecks or fault points, and improve system maintainability and operational efficiency.

[0101] 6. Dynamic environmental adaptation

[0102] The dynamic node management and authentication module performs the following operations:

[0103] Beacon listening and capture: A service runs on each fixed or mobile node to continuously listen to a pre-agreed underwater acoustic frequency band (such as a specific LF band).

[0104] Signal Decoding and Authentication: Upon capturing an authentication beacon signal conforming to the protocol format (containing an encrypted node ID and a hardware configuration description in JSON format), the signal is decoded and its characteristics are analyzed (to confirm that it originates from a trusted sound source). Subsequently, the decoded node ID is checked against a pre-issued whitelist.

[0105] Resource Registration: After successful authentication, the management module instructs the new node to complete a standard network connectivity and benchmark performance test. Upon successful testing, the node's detailed information (IP address, hardware configuration list) is sent to the resource status monitoring module of the scheduling platform, dynamically updating the global resource view.

[0106] It enables plug-and-play functionality for mobile heterogeneous nodes, allowing the system to expand flexibly and adapt to dynamic combat or detection formations such as naval formations and underwater clusters, significantly improving the system's resilience and scenario adaptability.

[0107] The mobility and link quality awareness module performs the following operations:

[0108] Information Acquisition: This module obtains the ship's position from the ship's navigation system, periodically sends detection signals or receives beacons from neighboring ships via the underwater acoustic communication device, and estimates the distance between nodes and relative speed; at the same time, it monitors the signal-to-noise ratio (SNR), bit error rate, and signal strength (RSSI) of the underwater acoustic channel.

[0109] Quality assessment and prediction: Taking into account factors such as distance, relative speed, SNR, RSSI, etc., a link quality index (LQI) between 0 and 1 is calculated through a weighted model, and its short-term trend is predicted based on historical data (e.g., due to ship maneuvering, the LQI will decrease after 30 seconds).

[0110] Scheduling optimization: When selecting deployment nodes for containers, the process parsing and scheduling module, in addition to considering hardware resources, will also query this module to prioritize nodes whose current and predicted LQI with the data source or collaborating containers are both above the threshold (e.g., 0.7) and whose communication latency is below the requirement (e.g., 50ms). During task execution, if it is predicted that the quality of a critical link will deteriorate significantly, the scheduling platform can proactively trigger container migration.

[0111] Transforming network conditions from a black box into an explicit input for scheduling decisions enables task scheduling to be aware of the environment. This effectively avoids communication interruptions or performance drops caused by node movement or channel fading, thus ensuring the continuous, stable, and high-performance operation of distributed processing tasks in dynamic and uncertain underwater network environments.

[0112] 7. Multi-layered security isolation

[0113] Containerized security sandbox module: At container startup, through Linux kernel security modules (such as AppArmor, SELinux) or container runtime security configurations, strict resource access control lists (such as allowing access to specific directories) and system call whitelists (such as prohibiting certain high-risk system calls) are defined for each algorithm container instance. During runtime, a lightweight auditing process logs abnormal behavior of the container.

[0114] Dynamic security classification and access control:

[0115] Tagging: When data enters the system from the source (such as a certain type of sonar) or after being processed by a specific algorithm (such as a classification and recognition algorithm), the dynamic security level tagging module will automatically add a security level tag (such as "public", "internal" or "secret") to the data packet according to preset rules (such as data source type and processing result confidence).

[0116] Control: The security-level-based access control gateway acts as a security checkpoint on the data channel. When a container instance attempts to send or receive data to another container, the gateway intercepts the data packet, checks its security level label, and verifies the authorized security level range of the target container. Only data flows that conform to the security level backward compatibility or same-level rules are allowed to pass; otherwise, they are blocked and logged in the security log.

[0117] A robust security defense is built within an open architecture. The security sandbox prevents malicious or flawed algorithms from damaging or interfering with the host system or other tasks. Dynamic security level access control ensures data flows within the system with the least privilege, effectively preventing unauthorized access and leakage of highly sensitive information.

[0118] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made to it in form and detail without departing from the scope defined by the claims of the present invention.

Claims

1. An open architecture system for an underwater acoustic signal processor, characterized in that, include: The hardware resource pool includes multiple computing nodes containing dedicated hardware units for underwater acoustic signal processing. These computing nodes are connected through an interconnection network to form a set of computing resources for dynamic resource allocation. The standardized algorithm container group includes multiple independently packaged containerized underwater acoustic signal processing algorithm units. Each algorithm unit packages the preset algorithm code, dependency libraries and configuration files into a uniform format image, and declares the type and number of dedicated hardware units required for its operation. The visualization processing link building module is used by users to connect and configure parameters of multiple algorithm units in a standardized algorithm container group according to preset processing logic, define a complete underwater acoustic signal processing flow, and generate a flow description file that can be executed by a containerized deployment and scheduling platform. The containerized deployment and scheduling platform is used to obtain process description files from the visual processing link building module, and instantiate and schedule algorithm containers to the matching dedicated hardware units for execution based on the needs of real-time processing tasks and the availability of hardware resource pools; it includes a mobility and link quality awareness module, which is used to continuously evaluate the relative position changes between the computing nodes of each ship, the quality of the underwater acoustic channel and the status of the radio frequency link, obtain the link quality index and predict its changing trend. When performing resource matching, the process parsing and scheduling module prioritizes the deployment of containers on computing nodes whose predicted link quality index with the data source node or upstream and downstream processing nodes is higher than a preset value and whose communication latency is lower than a preset threshold. The data processing and communication plane establishes data flow channels between various algorithm container instances to transmit underwater acoustic data, including raw acoustic data, intermediate processing results, and final situational information. The dynamic node management and authentication module is used to detect and identify newly added ship computing nodes through underwater acoustic signals, verify their identity and capabilities, and complete their registration or deregistration in the hardware resource pool. Specifically, the dynamic node management and authentication module performs the following operations: listening to a preset underwater acoustic signaling frequency band, capturing the underwater acoustic authentication beacon signal broadcast by the new node, which contains its node ID and hardware configuration; decoding and analyzing the captured underwater acoustic authentication beacon signal, and verifying whether the node ID exists in the preset trusted node list. If authentication is successful, the network access test of the new node is completed, and its hardware resource configuration information is updated in the global resource view of the containerized deployment and scheduling platform so that it can be used for task scheduling.

2. The open architecture system for an underwater acoustic signal processor according to claim 1, characterized in that: The containerized deployment and scheduling platform includes: The resource status monitoring module continuously collects load, utilization and health status data of each dedicated hardware unit in the hardware resource pool to form a global resource view; The process parsing and scheduling module receives the process description file generated by the visualization processing link construction module, parses out the algorithm container dependencies, resource requirements and data flow contained therein, and, based on the global resource view, uses a comprehensive evaluation algorithm to allocate corresponding computing nodes and dedicated hardware units to each algorithm container instance. The container lifecycle management module executes the instantiation, startup, shutdown, and resource reclamation operations of algorithm containers on the target computing node specified by the process parsing and scheduling module, and provides runtime status monitoring and fault recovery functions for running container instances.

3. The open architecture system for an underwater acoustic signal processor according to claim 2, characterized in that: The process parsing and scheduling module performs resource matching and scheduling in the following manner: When parsing the process description file, identify the dedicated hardware unit requirements declared for each algorithm container; In the global resource view, a set of candidate computing nodes that currently meet the dedicated hardware requirements and have sufficient available resources is selected; If there are multiple candidate nodes, the optimal node will be selected for deployment based on a comprehensive evaluation of each node's real-time load, data communication overhead with upstream and downstream containers, and historical task execution efficiency. If no node currently meets the mandatory hardware requirements of all containers, a resource shortage alarm is returned to the visualization processing link building module, or an alternative algorithm is executed on the general computing unit according to a preset strategy.

4. The open architecture system for an underwater acoustic signal processor according to claim 1, characterized in that: The data processing and communication plane performs the following operations: Based on shared memory and remote direct data access (RDMA) technology, cross-process or cross-node data channels are established between algorithm container instances located on the same node or different nodes. The data distribution model adopts a publish-subscribe approach. Algorithm container instances act as data producers, publishing results to specified topics. Algorithm container instances that need this data as input subscribe to the corresponding topics, and the communication plane completes the real-time forwarding and routing of data. Configure differentiated transmission quality of service strategies for different types of underwater acoustic data streams, including ensuring that the throughput of the broadband stream data after beamforming is greater than a preset throughput threshold, and ensuring that the latency of the target tracking command data is less than a preset latency threshold.

5. The open architecture system for an underwater acoustic signal processor according to claim 1, characterized in that: The visualization processing link construction module includes: An interactive orchestration canvas graphically displays selectable algorithm container units, which can be used to build processing chains from raw data input to final output through drag-and-drop and connection operations; The link logic verification unit automatically checks whether the input and output data formats, sampling rates, and number of channels between adjacent algorithm container units in the link are compatible during the orchestration process, and provides modification prompts. Run the simulation unit to inject historical or simulation data, simulate the execution process of the constructed processing link in an offline environment, and preview the data processing results and performance indicators of preset key links.

6. The open architecture system for an underwater acoustic signal processor according to claim 1, characterized in that: It also includes a unified maintenance and management plane, which includes: The algorithm image repository service is used for centralized storage, version management, and distribution of all algorithm images of the standardized algorithm container group. It supports users to submit new version images and records each update log. The system configuration and monitoring center is used to centrally configure the network parameters of the hardware resource pool, the strategies of the data processing and communication plane, and monitor the overall operating status, performance indicators and alarm information of the entire system and each algorithm container.

7. An open architecture system for an underwater acoustic signal processor according to claim 1, characterized in that: It also includes a multi-layered security isolation and access control plane, which includes: a containerized security sandbox module, used to configure resource access control lists and system call whitelists for each algorithm container instance, and to perform real-time monitoring and auditing during the runtime of the algorithm container instance; a dynamic security level labeling module, used to dynamically label the security level of data packets flowing through the container based on data source attributes or processing stages; and a security level label-based access control gateway, which is located between the data channels of the container instances and controls the forwarding and blocking of data flows based on the security level labels and the permission level of the target container.

Images