A method for improving channel traffic capacity based on underwater map

Abstract

The application discloses a kind of based on underwater map improves the method of channel traffic volume, belong to water transport technical field, including the following steps: S1 constructs underwater high-precision map, deploys unmanned ship, underwater robot, utilizes multibeam sounding technology to collect underwater topographic data to construct high-precision underwater map;S2 dynamic channel planning model is established, using big data analysis and machine learning algorithm, collect the hydrological data of channel different time and place and ship traffic data, generate dynamic channel use scheme;S3 integration ship navigation and management system, develop ship navigation terminal equipment, integrate dynamic channel planning model with the monitoring system of channel management department.This application optimizes channel resource utilization, realizes intelligent dynamic planning, improves ship traffic efficiency, navigation autonomy, navigation safety and system stability, reliability, enhances the scientific nature of channel management, strengthens the coordination of ship and management department.

Description

Technical Field

[0001] This invention relates to the field of water transport technology, specifically to a method for improving waterway traffic capacity based on underwater maps. Background Technology

[0002] With the rapid development of the global economy, water transport, as an important mode of cargo transportation, is experiencing ever-increasing demand. Waterways, as key channels in water transport, bear a large volume of vessel traffic. However, many waterways currently face severe congestion problems, which greatly limit water transport efficiency, increase transportation costs, and even affect the development of related industries.

[0003] In actual navigation, a lack of understanding of underwater topography is a significant factor contributing to channel congestion. Underwater topography is complex and varied, featuring reefs, shoals, gullies, and other landforms. For vessels, accurately understanding underwater topography is crucial for ensuring navigational safety. Small vessels, in particular, due to their smaller tonnage and relatively weaker wave resistance, often choose to navigate in the middle of the channel when faced with unknown underwater terrain, for safety reasons. This is because the middle of the channel is generally considered a relatively safe area, with deeper and more stable water, reducing the risk of running aground or striking reefs.

[0004] However, the concentration of small vessels in the middle of the channel has brought many adverse effects. Channel resources are inherently limited, and the significant occupation of the valuable middle section by small vessels obstructs the passage of larger vessels. Large vessels, with their larger size and cargo capacity, require wider and deeper channel space for navigation. When the middle of the channel is occupied by small vessels, large vessels are forced to reduce their speed or wait for smaller vessels to give way. This not only reduces the transport efficiency of large vessels but may also lead to traffic congestion, further exacerbating the congestion in the channel.

[0005] Furthermore, the hydrological conditions of waterways change with the seasons and times of day, including fluctuations in water levels and changes in current speed and direction. These changes affect underwater topography, and previously safe waterway areas may become unsuitable for navigation at times. However, existing waterway management methods often fail to fully consider these dynamic factors and lack effective means to rationally plan waterway use according to different seasons and times.

[0006] Currently, while some traditional underwater topographic survey methods exist, such as single-beam echo sounders and multi-beam echo sounders, these methods have certain limitations. Single-beam echo sounders can only measure the water depth directly below a vessel, failing to acquire large-area underwater topographic information and thus exhibiting low efficiency. While multi-beam echo sounders can acquire underwater topographic data within a certain range, they still have shortcomings in data processing, real-time updates, and integration with waterway management systems. Moreover, the data acquired by these traditional methods is often not fully and effectively utilized, failing to form an organic integration with waterway traffic management and thus unable to fundamentally solve the problem of waterway congestion. Therefore, developing a technology based on underwater maps that can rationally allocate waterway usage according to actual conditions is an urgent practical need for improving waterway traffic capacity. Summary of the Invention

[0007] The purpose of this invention is to provide a method for improving waterway traffic capacity based on underwater maps, which solves the problems of congestion caused by small vessels irrationally occupying waterway resources due to a lack of understanding of underwater topography and the lack of dynamic planning for waterway use.

[0008] Technical solution: This invention provides a method for improving waterway traffic capacity based on underwater maps, comprising the following steps:

[0009] S1 constructs high-precision underwater maps by deploying unmanned vessels and underwater robots, using multibeam echo sounding technology to collect underwater topographic data, and using satellite positioning and inertial navigation systems to ensure the accuracy of data collection locations. The raw data is then processed to construct high-precision underwater maps.

[0010] S2 establishes a dynamic waterway planning model, collects hydrological data of different seasons and times of the waterway as well as historical ship passage data, and uses big data analysis and machine learning algorithms to generate dynamic waterway usage plans based on real-time hydrological conditions and ship passage needs.

[0011] The S3 integrated ship navigation and management system includes the development of ship navigation terminal equipment, which integrates underwater map display, dynamic waterway planning information reception, and ship position positioning functions; it also includes the integration of dynamic waterway planning models with the monitoring system of waterway management departments to realize data interaction and collaborative management between ships and management departments.

[0012] Furthermore, in the aforementioned method for improving channel traffic capacity based on underwater maps, step S1 includes the following steps:

[0013] S11 data acquisition involves deploying unmanned surface vessels (USVs) and underwater robots, equipped with multibeam echo sounders and sonar equipment. USVs efficiently conduct large-scale data acquisition in wide waterways, while underwater robots excel at penetrating narrow, complex, or shallow terrain areas for detailed exploration.

[0014] The S12 data processing and map building process first denoises the collected data, preserving the effective low-frequency components. Then, it filters the data, adjusting the filtering parameters in real-time based on dynamic data changes to improve stability and accuracy. Next, a feature-matching-based stitching algorithm is used to stitch the data together, extracting feature points from overlapping areas of data collected from different devices. A feature matching matrix is ​​constructed to align feature points from different datasets, achieving seamless stitching and building an underwater terrain model. Finally, the underwater terrain model is converted into a vector or raster format underwater map supported by a Geographic Information System (GIS), stored in a database, and a data retrieval mechanism based on spatial indexes (such as R-tree indexes) is established to enable fast and efficient map data querying and retrieval, meeting the application needs of real-time waterway management and ship navigation.

[0015] Furthermore, in the aforementioned method for improving channel traffic capacity based on underwater maps, step S2 includes the following steps:

[0016] S21 Multivariate data collection and integration: Working closely with hydrological monitoring departments, detailed hydrological data for different seasons and times of the waterway is obtained. Simultaneously, historical vessel traffic data is extracted from the waterway management department's database, classifying vessel sizes and types, and using size parameters to determine the required navigation space for vessels in the waterway. A time synchronization mechanism is established to link hydrological data and vessel traffic data within the same time interval, forming a complete data sample set D, providing a rich and accurate data foundation for subsequent model training. S22 Machine learning model construction and training: A dynamic waterway planning model is constructed using the Long Short-Term Memory (LSTM) algorithm in deep learning. The data sample set D is input into the model for training, outputting recommended navigation routes for different vessel types under current hydrological conditions. The algorithm is continuously optimized to adjust the model's weights and biases, enabling the model to accurately predict reasonable navigation routes based on the input hydrological and vessel information.

[0017] The S23 model is updated and optimized in real time. At fixed time intervals, the latest hydrological data and vessel traffic data are obtained from hydrological monitoring departments and waterway management departments, and the new data is added to the data sample set D to form an expanded dataset D. new Incremental training is performed on the already trained LSTM model. During incremental training, the model structure remains unchanged, and only the model weights and biases are fine-tuned. At the same time, model performance evaluation metrics are introduced, and the training parameters of the model are dynamically adjusted according to the evaluation results to ensure that the model always maintains the best performance state and provides accurate and reliable support for real-time channel planning.

[0018] Furthermore, in the aforementioned method for improving channel traffic capacity based on underwater maps, step S3 includes the following steps:

[0019] The development of the S31 ship navigation terminal equipment is aimed at the complex and harsh navigation environment of ships. The core hardware processor of the terminal equipment is selected with a chip with high performance computing power and low power consumption. The screen adopts a high brightness and high contrast display. The positioning module integrates the Global Navigation Satellite System (GNSS) and Ultra-Wideband (UWB) positioning technology. The communication module supports multiple communication protocols. A navigation software system with multiple modules working together has been developed.

[0020] The S32 waterway management system is integrated with the model, and an information receiving module is set up to parse the navigation instructions issued by the waterway management system. The instructions contain a series of coordinate points for the recommended route. By comparing with the ship's current position coordinates (x ship ,y ship In contrast, the path planning algorithm is used to calculate the optimal navigation path from the current position to the recommended route. The positioning module obtains the ship's position in real time and continuously tracks the ship's position by combining the trajectory estimation algorithm. Through continuous iteration, high-precision trajectory estimation is achieved to compensate for the position monitoring when the positioning signal is briefly lost.

[0021] The S33 system operates and is maintained through real-time two-way data interaction between the ship navigation terminal and the waterway management system. The waterway management system generates navigation instructions based on a dynamic waterway planning model and sends them to the ship navigation terminal via the communication network. Upon receiving the instructions, the ship displays them on its navigation software interface and plans its navigation path accordingly. Simultaneously, the ship uploads its own navigation status data to the waterway management system in real time. The waterway management system monitors ship navigation based on this data, performs global optimization management of waterway traffic, and establishes a system fault diagnosis and fault tolerance mechanism. When a fault occurs, fault tree analysis (FTA) technology is used to quickly locate the fault point. For some non-critical faults, the system employs fault-tolerant technology to ensure that the system can maintain basic operation even under certain fault conditions, guaranteeing the safety and continuity of waterway passage.

[0022] Furthermore, in the above-mentioned method for improving waterway traffic capacity based on underwater maps, in step S11, the unmanned vessel and underwater robot plan the navigation path according to the shape, width, depth of the waterway and known terrain features. The multibeam echo sounder and sonar equipment work together to collect data, and the collected data can be transmitted in real time and backed up locally.

[0023] Furthermore, in the aforementioned method for improving channel traffic capacity based on underwater maps, in step S11, the multibeam echo sounder uses acoustic reflection ranging, and the emitted acoustic beam forms a certain opening angle θ with the horizontal plane. Assuming the speed of sound in water is v, it is corrected using an empirical formula:

[0024] v = 1449.2 + 4.6T - 0.055T 2 +0.00029T3 +(1.34-0.01T)(S-35)+0.016D

[0025] Where T is water temperature in °C; S is salinity; and D is depth in meters.

[0026] After a multibeam echo sounder emits sound waves, it receives the signals reflected back from the underwater topography and measures the round-trip propagation time t of the sound waves. The water depth h at the corresponding measuring point can then be calculated using the formula... Calculations show that the multibeam echo sounder continuously emits sound beams at high frequencies (generally between 100Hz and 500Hz) to form a dense array of measurement points, enabling high-resolution scanning of the underwater topography of the waterway. Simultaneously, the sonar equipment works in sync, transmitting broadband sound signals and receiving echo signals to construct an acoustic image of the underwater topography.

[0027] Furthermore, in the above-mentioned method for improving waterway traffic capacity based on underwater maps, in step S11, to ensure high-precision positioning of the collected data, a satellite positioning system (such as GPS or BeiDou) and an inertial navigation system are integrated. The satellite positioning system uses signals emitted by multiple satellites to determine the device position through trilateration, while the inertial navigation system measures the acceleration and angular velocity of the device based on accelerometers and gyroscopes, and realizes coordinate transformation through attitude matrix to track the device's position and attitude changes in space in real time.

[0028] Furthermore, in the above-mentioned method for improving waterway traffic capacity based on underwater maps, the hardware of the ship navigation terminal equipment in step S31 has waterproof, dustproof, and shockproof performance, and the software realizes the functions of underwater map display, dynamic waterway planning information reception, and ship position positioning.

[0029] Furthermore, in the aforementioned method for improving waterway traffic capacity based on underwater maps, step S32 also requires the establishment of an efficient communication network. This involves setting up communication base stations on shore and employing a hybrid communication mode combining cellular and satellite communication. In areas with good signal coverage along the waterway, ships communicate with the base stations via 4G / 5G cellular networks. In remote areas or areas with severe signal obstruction, communication is switched to satellite to ensure continuous data transmission. Additionally, ships establish an Ad-Hoc communication network. When multiple ships are within close proximity, an Ad-Hoc network automatically forms to exchange information and achieve efficient data forwarding between ships, ensuring timely information transmission and enhancing the collaborative navigation capabilities of ships within the waterway.

[0030] Furthermore, the above-mentioned method for improving channel traffic capacity based on underwater maps includes the following steps: Step S22 includes the following steps:

[0031] The S221 model receives the integrated data sample set D at the input layer, including water level H, water flow velocity V, water flow direction α, and ship type S. type (converted to vector form via one-hot encoding), ship length L ship Ship width W ship As the input feature vector X(t), inside the LSTM network, the input vector X(t) at each time step t interacts with the hidden state h(t-1) and the cell state C(t-1) of the previous time step;

[0032] The S222 model's LSTM unit controls the flow of information through an input gate i(t), a forget gate f(t), and an output gate o(t). The input gate calculation formula is i(t) = σ(W). ix X(t)+W ih h(t-1)+b i The forgetting gate calculation formula is f(t) = σ(W). fx X(t)+W fh h(t-1)+b f The output gate calculation formula is o(t) = σ(W). ox X(t)+W oh h(t-1)+b o Through the regulation of these gates, the cell state C(t) is updated to C(t) = f(t)e^(-C(t-1)) + i(t)e^(tanh(W)). cx X(t)+W ch h(t-1)+b c The new hidden state is h(t) = o(t)e tanh(C(t)).

[0033] Where σ is the sigmoid activation function, W ix W ih W fx W fh W ox W oh Let b be the weight matrix. i ,b f ,b o Let e ​​be the bias vector, and e denotes element-wise multiplication.

[0034] The output layer of the S223 model outputs recommended navigation routes for different ship types under current hydrological conditions based on the hidden states processed by the LSTM network. The navigation routes are represented by a series of discrete coordinate points, and a regression algorithm maps the hidden states to coordinate values. Let the output navigation route coordinate points be (x... out (t),y out(t)) The model is trained by minimizing the mean squared error loss function between the predicted route and the historical best route. The model weights and biases are continuously adjusted by optimization algorithms such as stochastic gradient descent, so that the model can accurately predict reasonable navigation routes based on the input hydrological and ship information.

[0035] As can be seen from the above technical solution, the present invention has the following beneficial effects:

[0036] 1. Enhanced navigation safety: Utilizing advanced multibeam echo sounding technology and flexible data acquisition carriers such as unmanned vessels and underwater robots, the constructed high-precision underwater maps provide detailed representations of various underwater terrain features, including reefs, shoals, and gullies. Ship operators can clearly view the underwater terrain around their vessels through navigation terminals, plan safe routes in advance, effectively avoid accidents such as running aground or hitting reefs, and greatly improve the navigation safety of ships in complex waterway environments.

[0037] 2. Optimize waterway resource utilization: Accurate underwater maps provide waterway management departments with a precise data foundation, enabling them to plan waterways more scientifically. For example, based on the actual underwater topography, navigation areas for different types of vessels can be rationally divided, and waterway resources can be allocated in a refined manner. This avoids resource waste caused by small vessels irrationally occupying the middle positions of the waterway, creating conditions for increasing waterway throughput. 3. Adapt to dynamic hydrological changes: Combining satellite positioning systems, inertial navigation systems, and real-time hydrological data, underwater maps can reflect the dynamic changes in underwater topography caused by factors such as water level fluctuations and current changes in different seasons and times. This allows vessels and waterway management departments to adjust navigation and management strategies in a timely manner, adapting to complex and ever-changing hydrological environments and ensuring the continuous safety and efficient passage of waterways.

[0038] 4. Achieve intelligent dynamic planning: The dynamic waterway planning model established by using big data analysis and machine learning algorithms can fully consider the real-time hydrological conditions of the waterway and the passage needs of ships. Compared with the traditional fixed waterway planning method, this model can automatically generate the optimal waterway usage plan according to the actual situation of different seasons and times, realizing the intelligence and dynamism of waterway planning and greatly improving the waterway capacity.

[0039] 5. Improved vessel traffic efficiency: The model provides personalized recommended navigation routes for different types of vessels, allowing ship operators to navigate more efficiently based on the planning information received from the navigation terminal; large vessels can quickly find the most suitable wide and deep channel areas for their navigation, while small vessels can also make reasonable use of the channel space near the shore while ensuring safety; it reduces mutual interference and waiting time between vessels, significantly improves vessel traffic efficiency, and reduces transportation costs.

[0040] 6. Enhance the scientific nature of waterway management: Based on the scheme generated by the dynamic waterway planning model, the waterway management department can more rationally schedule and manage ships in the waterway. By monitoring the navigation of ships in real time, the waterway usage strategy can be adjusted in a timely manner to avoid waterway congestion, thereby improving the scientific nature and accuracy of waterway management.

[0041] 7. Enhance ship navigation autonomy: Ship navigation terminal equipment integrates multiple functions, allowing ship operators to independently obtain the necessary navigation information and autonomously plan navigation routes, reducing reliance on external navigation support and improving the autonomy and flexibility of ship navigation.

[0042] 8. Strengthen collaboration between vessels and management departments: Real-time two-way data interaction between vessel navigation terminals and waterway management systems enables efficient collaboration between vessels and management departments. Vessels can receive navigation instructions issued by management departments in a timely manner and upload their own navigation status data to management departments, enabling management departments to keep abreast of the dynamics of all vessels in the waterway and carry out global optimization management, further improving the efficiency and safety of waterway traffic.

[0043] 9. Improve system stability and reliability: The fault diagnosis and fault tolerance mechanism established during system integration effectively ensures the stable operation of the ship navigation and management system; when a fault occurs, the system can quickly diagnose the cause of the fault and take corresponding fault tolerance measures to ensure that basic functions can still be maintained under certain fault conditions, reduce the impact of system faults on waterway passage, and improve the reliability of the entire system. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the process of constructing an underwater high-precision map in step S1 of the present invention;

[0045] Figure 2 This is a schematic diagram of the process of establishing a dynamic waterway planning model in S2 of the present invention;

[0046] Figure 3 This is a schematic diagram of the S3 integrated ship navigation and management system of the present invention. Detailed Implementation

[0047] Example 1

[0048] like Figure 1-3 The method shown is to improve waterway traffic based on underwater maps, including: S1 constructing a high-precision underwater map, deploying unmanned vessels and underwater robots, collecting underwater topographic data using multibeam echo sounding technology, ensuring the accuracy of data collection location through satellite positioning system and inertial navigation system, and processing the collected raw data to construct a high-precision underwater map;

[0049] S2 establishes a dynamic waterway planning model, collects hydrological data of different seasons and times of the waterway as well as historical ship passage data, and uses big data analysis and machine learning algorithms to generate dynamic waterway usage plans based on real-time hydrological conditions and ship passage needs.

[0050] The S3 integrated ship navigation and management system includes the development of ship navigation terminal equipment, which integrates underwater map display, dynamic waterway planning information reception, and ship position positioning functions; it also includes the integration of dynamic waterway planning models with the monitoring system of waterway management departments to realize data interaction and collaborative management between ships and management departments.

[0051] like Figure 1 The method shown is for improving channel traffic capacity based on underwater maps. Step S1 includes the following steps:

[0052] The S11 data acquisition system deploys unmanned surface vessels (USVs) and underwater robots, equipped with multibeam echo sounders and sonar. The USVs are selected for their high performance and flexible beam angle adjustment. The USVs efficiently collect data over a wide area in wide waterways, while the underwater robots excel at exploring narrow, complex, or shallow terrain areas for detailed exploration.

[0053] The S12 data processing and map building process first denoises the collected data, preserving the effective low-frequency components. Then, it filters the data, adjusting the filtering parameters in real-time based on dynamic data changes to improve stability and accuracy. Next, a feature-matching-based stitching algorithm is used to stitch the data together, extracting feature points from overlapping areas of data collected from different devices. A feature matching matrix is ​​constructed to align feature points from different datasets, achieving seamless stitching and building an underwater terrain model. Finally, the underwater terrain model is converted into a vector or raster format underwater map supported by a Geographic Information System (GIS), stored in a database, and a data retrieval mechanism based on spatial indexes (such as R-tree indexes) is established to enable fast and efficient map data querying and retrieval, meeting the application needs of real-time waterway management and ship navigation.

[0054] like Figure 2 The method shown in the figure for improving channel traffic capacity based on underwater maps includes the following steps in step S2:

[0055] The S21 multi-source data acquisition and integration system collaborates closely with hydrological monitoring departments to obtain detailed hydrological data for different seasons and times in the waterway. Water level data is denoted as H(t), where t represents time, accurately recording water level changes at various times in the waterway. These changes directly affect the safe water depth for vessel navigation. Current velocity V(t), measured using current meters and other equipment, reflects the dynamic effect of the current on vessel navigation, measured in m / s. Current direction α(t), expressed as an angle, determines the direction of the current that vessels need to overcome or utilize, crucial for planning vessel navigation paths. Simultaneously, historical vessel traffic data is extracted from the waterway management department's database, including information on vessel type, size, travel time, and trajectory. This data is cleaned to remove invalid or erroneous records, and the valid data is organized according to time series. Vessel size is categorized into small vessels (S), medium vessels (M), and large vessels (L), with vessel dimensions including length L. ship Width W ship The required navigation space for a ship in a waterway is determined by its dimensional parameters, and the navigation time t travel Record the specific times when the ship passed through the waterway, and the navigation route is represented by a series of geographical coordinates (x, y, y). i ,y i Let i = 1, 2, ..., n, to depict the actual navigation trajectory of a ship. A time synchronization mechanism is established to link hydrological data with ship passage data within the same time interval. Within the time interval [t1, t2], the corresponding water level H([t1, t2]), current velocity V([t1, t2]), current direction α([t1, t2]), and navigation information of various types of ships passing through during this period are integrated to form a complete data sample set D = {(H, V, α, S...}. type ,L ship W ship ,t travel ,(x i ,y i This provides a rich and accurate data foundation for subsequent model training.

[0056] The model building and training of S22 machine learning uses the Long Short-Term Memory (LSTM) network algorithm in deep learning to build a dynamic waterway planning model. The data sample set D is input into the model for training, and the output is a recommended navigation route for different ship types under the current hydrological conditions. The algorithm is optimized to continuously adjust the model's weights and biases, so that the model can accurately predict reasonable navigation routes based on the input hydrological and ship information.

[0057] The S23 model is updated and optimized in real time. At fixed time intervals, the latest hydrological data and vessel traffic data are obtained from hydrological monitoring departments and waterway management departments, and the new data is added to the data sample set D to form an expanded dataset D. newIncremental training is performed on the already trained LSTM model. During incremental training, the model structure remains unchanged, and only the model weights and biases are fine-tuned. At the same time, model performance evaluation metrics are introduced, and the training parameters of the model are dynamically adjusted according to the evaluation results to ensure that the model always maintains the best performance state and provides accurate and reliable support for real-time channel planning.

[0058] like Figure 3 The method shown in the figure for improving channel traffic capacity based on underwater maps includes the following steps in step S3:

[0059] The development of the S31 ship navigation terminal equipment is aimed at the complex and harsh navigation environment of ships. The core hardware processor of the terminal equipment is selected with a chip with high performance computing power and low power consumption. The screen adopts a high brightness and high contrast display. The positioning module integrates the Global Navigation Satellite System (GNSS) and Ultra-Wideband (UWB) positioning technology. The communication module supports multiple communication protocols. A navigation software system with multiple modules working together has been developed.

[0060] The S32 waterway management system is integrated with the model, and an information receiving module is set up to parse the navigation instructions issued by the waterway management system. The instructions contain a series of coordinate points for the recommended route. By comparing with the ship's current position coordinates (x ship ,y ship In comparison, the optimal navigation path from the ship's current position to the recommended route is calculated using a path planning algorithm. The cost function of the planned path is F = G + H, where G is the actual cost from the starting point to the current node, and H is the estimated cost from the current node to the target node. The positioning module acquires the ship's position in real time and continuously tracks the ship's position using a trajectory extrapolation algorithm. Let the ship's initial position be (x0, y0), the heading angle be θ, and the speed be v. Within the time interval Δt, the ship's position is updated to x = x0 + vΔtcosθ and y = y0 + vΔtsinθ. High-precision trajectory extrapolation is achieved through continuous iteration, compensating for position monitoring when the positioning signal is briefly lost.

[0061] The S33 system operates and is maintained by enabling real-time two-way data exchange between the ship's navigation terminal and the waterway management system. After generating navigation instructions based on the dynamic waterway planning model, the waterway management system sends them to the ship's navigation terminal via the communication network. Upon receiving the instructions, the ship displays them on its navigation software interface and plans its navigation path accordingly. Simultaneously, the ship uploads its own navigation status data (such as speed, heading, position, and equipment status) to the waterway management system in real time. Let the amount of data uploaded by the ship be D. up The upload frequency is f up Then the data upload rate F per unit time up =D up f upThe waterway management system monitors ship navigation in real time based on this data, optimizes waterway traffic globally, and collects feedback from ship operators and waterway management personnel. It promptly optimizes and improves systems to address issues such as communication delays and data display errors. A system maintenance team is established to regularly maintain ship navigation terminal equipment, the waterway management system, and the dynamic waterway planning model. A system fault diagnosis and fault tolerance mechanism is in place. When a fault occurs, the system promptly triggers a diagnostic process by monitoring key indicators (such as communication interruptions and data anomalies). Using techniques such as fault tree analysis (FTA), the system reverse-engineers the cause of the fault from the symptoms, quickly locating the fault point. For some non-critical faults, the system employs fault-tolerant techniques, such as data redundancy backup and recovery, and backup communication link switching, to ensure that the system can maintain basic operation even under certain fault conditions, guaranteeing the safety and continuity of waterway traffic.

[0062] Example 2

[0063] Based on Example 1, in this example, in step S11, the unmanned vessel and underwater robot plan a navigation path based on the shape, width, depth of the channel and known terrain features. The multibeam echo sounder and sonar equipment work together to collect data, and the collected data can be transmitted in real time and backed up locally.

[0064] In this embodiment, the multibeam echo sounder in step S11 is based on acoustic wave reflection ranging. The emitted acoustic beam forms a certain opening angle θ with the horizontal plane. Assuming the speed of sound in water is v, it is corrected using an empirical formula:

[0065] v = 1449.2 + 4.6T - 0.055T 2 +0.00029T 3 +(1.34-0.01T)(S-35)+0.016D

[0066] Where T is water temperature in °C; S is salinity; and D is depth in meters.

[0067] After a multibeam echo sounder emits sound waves, it receives the signals reflected back from the underwater topography and measures the round-trip propagation time t of the sound waves. The water depth h at the corresponding measuring point can then be calculated using the formula... Calculations show that the multibeam echo sounder continuously emits acoustic beams at high frequencies (typically 100Hz-500Hz) to form a dense array of measurement points, achieving high-resolution scanning of the underwater topography of the waterway. Simultaneously, the sonar equipment operates concurrently. The sonar transmits broadband acoustic signals and receives information such as the amplitude, phase, and time delay of the echo signals to construct an acoustic image of the underwater topography. Using signal processing algorithms, the received raw sonar signal S(t) is converted into image grayscale values ​​I(x,y) reflecting the underwater topographic features. The conversion relationship can be expressed as follows: Where t1 and t2 are the signal acquisition time periods, and g is the image generation function based on signal characteristics and target position τ(x,y). By analyzing and processing the echo signals at different locations, detailed underwater terrain texture and contour information are generated.

[0068] In this embodiment, in step S11, to ensure high-precision positioning of the collected data, a satellite positioning system (such as GPS or BeiDou) and an inertial navigation system are integrated. The satellite positioning system uses signals emitted by multiple satellites to determine the device's position through trilateration. The inertial navigation system, based on accelerometers and gyroscopes, measures the acceleration and angular velocity of the device and performs coordinate transformation through an attitude matrix, tracking the device's position and attitude changes in space in real time. The two positioning and navigation systems complement each other. Even in complex environments such as when satellite signals are blocked, the inertial navigation system can still ensure the accurate recording of the data acquisition device's position and attitude information, ensuring high-precision positioning of the collected data.

[0069] In this embodiment, GPS is used as an example. Let the position coordinates of satellite i be (x i ,y i ,z i The time it takes for the device to receive satellite signals is t. i Considering the signal propagation delay Δt, the device position coordinates (x, y, z) satisfy the equation (xx i ) 2 +(yy i ) 2 +(zz i ) 2 =(c(t) i -Δt)) 2 Where c is the speed of light. Accurate three-dimensional position information is obtained by simultaneously receiving signals from at least four satellites and solving a series of equations.

[0070] In this embodiment, the inertial navigation system measures the acceleration α of the carrier in three axes using accelerometers based on the measured acceleration and angular velocity. x a y a z After two integrations, the velocity v and displacement s are obtained. For example, in the x-axis direction, the velocity v... Displacement The gyroscope measures the angular velocity ω of the carrier. x ω y ω z Used to calculate the attitude angle θ of the carrier. x θ y θ z Coordinate transformation is achieved through attitude matrix, which tracks the position and attitude changes of the device in space in real time.

[0071] In this embodiment, the denoising process in the S12 data processing and map construction employs a wavelet transform denoising algorithm, a dedicated data processing algorithm. First, denoising is performed using a wavelet transform denoising algorithm. The original data sequence f(t) is decomposed into wavelet coefficients W at different frequency scales. j,k By setting a threshold λ, wavelet coefficients are filtered to remove those corresponding to noise, and then wavelet reconstruction is performed to obtain the denoised data f. d (t). In the frequency domain, noise signals are mainly distributed in the high-frequency band. Thresholding can effectively filter out noise while retaining the effective low-frequency components in the data.

[0072] In this embodiment, the filtering process in S12 data processing and map construction adopts the adaptive Kalman filter algorithm, and the system state equation is X. k =A k X k-1 +Q k-1 The observation equation is Z k =H k X k +R k , where X k Let A be the system state vector at time k. k Let Q be the state transition matrix. k-1 For process noise, Z k H is the observation vector. k Let R be the observation matrix. k To reduce observation noise, predicted values ​​are calculated iteratively. Predicting covariance Kalman gain and the optimal estimate The filtering parameters are adjusted in real time according to the dynamic changes of the data to improve the stability and accuracy of the data.

[0073] In this embodiment, during the data processing and map construction in S12, when constructing an underwater terrain model after data stitching, discrete measurement points are connected into triangular patches based on the Triangular Network Integer (TIN) algorithm. The model is then constructed according to the three-dimensional coordinates (x, y, z) of the measurement points. i ,y i ,z i A triangular topology is constructed so that each triangular facet can accurately reflect the local terrain features; the size and density of the triangles are dynamically adjusted according to the terrain undulations, and denser triangles are generated in areas with drastic terrain changes to improve the model accuracy.

[0074] Example 3

[0075] Based on the above embodiments, in this embodiment, step S22 uses the Long Short-Term Memory (LSTM) algorithm in deep learning to construct a dynamic waterway planning model. In a Python environment, the model structure is built using TensorFlow or PyTorch frameworks, setting up an input layer, multiple LSTM hidden layers, and an output layer. The input layer receives integrated hydrological data and ship feature data, and the output layer outputs recommended navigation routes for different ship types. The collected data is normalized, converting hydrological data (such as water level and current speed) and ship feature data (such as ship length and width) of different dimensions to the 0-1 range to facilitate model training. The processed data is divided into a 70% training set, a 20% validation set, and a 10% test set. During training, the training set data is input into the model, and the model's weights and biases are continuously adjusted using the backpropagation algorithm to minimize the mean square error between the model's predicted navigation route and the historical best route. A reasonable learning rate (e.g., 0.001) and number of training rounds (e.g., 500 rounds) are set, and GPU acceleration is used to improve training efficiency.

[0076] In this embodiment, step S22 includes the following steps:

[0077] The S221 model receives the integrated data sample set D at the input layer, including water level H, water flow velocity V, water flow direction α, and ship type S. type (converted to vector form via one-hot encoding), ship length L ship Ship width W ship As the input feature vector X(t), inside the LSTM network, the input vector X(t) at each time step t interacts with the hidden state h(t-1) and the cell state C(t-1) of the previous time step;

[0078] The S222 model's LSTM unit controls the flow of information through an input gate i(t), a forget gate f(t), and an output gate o(t). The input gate calculation formula is i(t) = σ(W). ix X(t)+W ih h(t-1)+b i The forgetting gate calculation formula is f(t) = σ(W). fx X(t)+W fh h(t-1)+b f The output gate calculation formula is o(t) = σ(W). ox X(t)+W oh h(t-1)+b o Through the regulation of these gates, the cell state C(t) is updated to C(t) = f(t)e^(-C(t-1)) + i(t)e^(tanh(W)). cx X(t)+Wch h(t-1)+b c The new hidden state is h(t) = o(t)e tanh(C(t)).

[0079] Where σ is the sigmoid activation function, W ix W ih W fx W fh W ox W oh Let b be the weight matrix. i ,b f ,b o Let e ​​be the bias vector, and e denotes element-wise multiplication.

[0080] The output layer of the S223 model outputs recommended navigation routes for different ship types under current hydrological conditions based on the hidden states processed by the LSTM network. The navigation routes are represented by a series of discrete coordinate points, and a regression algorithm maps the hidden states to coordinate values. Let the output navigation route coordinate points be (x... out (t),y out (t)) The model is trained by minimizing the mean squared error loss function between the predicted route and the historical best route. The model weights and biases are continuously adjusted by optimization algorithms such as stochastic gradient descent, so that the model can accurately predict reasonable navigation routes based on the input hydrological and ship information.

[0081] In this embodiment, during step S23, to avoid overfitting the model to new data during incremental training, regularization techniques, such as L2 regularization, are used by adding a regularization term to the loss function. Where λ is the regularization parameter and W is the model's weight matrix, the model can learn the latest channel change features through incremental training and continuously optimize its prediction performance.

[0082] In this embodiment, the model performance evaluation in step S23 uses mean absolute error (MAE) and root mean square error (RMSE) to compare and evaluate the model-predicted flight route with the actual optimal flight route:

[0083] The MAE calculation formula is: The formula for calculating RMSE is: in Represents the squared error term Based on the evaluation results, the training parameters of the model, such as the learning rate and regularization parameters, are dynamically adjusted to ensure that the model always maintains its optimal performance and provides accurate and reliable support for real-time channel planning.

[0084] In this embodiment, step S23 further includes deploying the real-time updated model to the waterway management system. When a vessel enters the waterway, the system inputs real-time hydrological data and vessel information into the model. The model quickly generates a waterway usage plan suitable for the current situation and transmits the plan to the vessel navigation terminal. Simultaneously, feedback from vessel drivers on the recommended route and actual navigation performance data are collected to further optimize the model.

[0085] Example 4

[0086] Based on the above embodiments, in this embodiment, the hardware of the ship navigation terminal equipment in step S31 has waterproof, dustproof and shockproof performance, and the software realizes the functions of underwater map display, dynamic waterway planning information reception and ship position positioning.

[0087] In this embodiment, the screen used in step S31 is a high-brightness, high-contrast display to ensure clear display under different lighting conditions, and its brightness adjustment range can reach L. min To L max (Unit: nits), which can be determined based on ambient light intensity I. env (Measured by an ambient light sensor, unit: lux), using the formula Real-time adjustment

[0088] Display brightness, where I env,max This is the maximum light intensity value that the ambient light sensor can detect.

[0089] In this embodiment, the positioning module in step S31 integrates Global Navigation Satellite System (GNSS) and Ultra-Wideband (UWB) positioning technologies. The GNSS module utilizes signals from multiple satellites to achieve high-precision positioning globally, calculating the position by receiving the time difference of satellite signal propagation. Let the position coordinates of satellite j be (x... j ,y j ,z j The signal propagation time is t. j Considering the signal propagation delay Δt j Then the terminal position coordinates (x, y, z) satisfy the equation (xx j ) 2 +(yy j ) 2 +(zz j ) 2 =(c(t) j -Δt j )) 2Where c is the speed of light. In complex waterways such as ports and narrow channels where satellite signals are easily blocked, UWB positioning technology serves as a supplement. UWB positioning base stations send pulse signals; after receiving the signal, the terminal calculates its relative position to the base station based on the time difference of arrival (TDOA). Let the distance between base station k and the terminal be d. k The distance between base stations is d km The precise location of the terminal is determined by simultaneously solving the measurement data from multiple base stations, as shown in the following formula:

[0090]

[0091] In this embodiment, the communication module in step S31 supports multiple communication protocols, including Very High Frequency (VHF) communication for short-range voice communication between ships and between ships and shore, with a communication distance r. VHF Under ideal conditions, it can be expressed by the formula Estimate, where R is the Earth's radius, h T h is the height of the transmitting antenna. R The antenna height is used for receiving signals; and 4G / 5G communication is used for high-speed data transmission to ensure that dynamic waterway planning information issued by the waterway management system can be received in real time and that the ship's own status data can be uploaded.

[0092] In this embodiment, the underwater map display in the navigation software system in step S31 adopts tile mapping technology, which divides the high-precision underwater map into tiles of different levels. The corresponding tiles are loaded in real time according to the ship's position and zoom level, achieving fast map display and smooth browsing. Let the map zoom level be z, and the total map size be 256. z ×256 z The formula for converting pixel coordinates (lat, lon) of a ship's location to map pixel coordinates (x, y) is as follows:

[0093]

[0094] In this embodiment, step S32 also requires the establishment of a high-efficiency communication network. Communication base stations are set up on shore, employing a hybrid communication mode combining cellular and satellite communication. In areas with good signal coverage along the waterway, ships communicate with base stations via 4G / 5G cellular networks, achieving data transmission rates up to R. 4G / 5G (Unit: Mbps), its rate is affected by factors such as signal strength and base station load, and can be calculated using Shannon's formula. Estimate the theoretical upper limit, where B is the channel bandwidth, S is the signal power, and N is the noise power; in remote areas or areas with severe signal obstruction, switch to satellite communication, and estimate the bit error rate P of the satellite communication link. eRelated to signal power and noise power spectral density, it can be calculated using relevant communication theory formulas to ensure the continuity of data transmission. Furthermore, ships establish self-organizing network (Ad-Hoc) communication. When multiple ships are within close range, they automatically form an Ad-Hoc network for information exchange, such as sharing navigation status and emergency warnings. In the Ad-Hoc network, the communication link quality Q between nodes (ships) is evaluated using parameters such as signal strength and bit error rate. Distributed routing algorithms (such as Dynamic Source Routing Protocol, DSR) are used to achieve efficient data forwarding between ships, ensuring timely information transmission between ships and improving the collaborative navigation capability of ships within the waterway.

[0095] Example 5

[0096] Based on the above embodiments, in this embodiment, step S11, data acquisition, involves calibrating the multibeam echo sounder in a simulated environment before the equipment is put into use. According to a standard water depth model, the transmission and reception parameters of the echo sounder are adjusted to minimize measurement errors. Comprehensive testing is conducted on the power, communication, and positioning systems of the unmanned surface vessel (USV) and underwater robot to ensure stable operation in complex waters. Using Geographic Information System (GIS) technology, combined with the shape, width, and depth of the waterway and known terrain features, detailed navigation paths are planned for the USV and underwater robot. In wide waterway areas with relatively simple terrain, the USV is planned to perform large-area scanning using a grid-like path, with the grid spacing set between 1 and 5 meters based on measurement accuracy requirements. For narrow, winding, or complex terrain areas, plan serpentine or spiral paths for underwater robots to ensure accurate detection of every terrain detail; precisely install multibeam echo sounders, sonar equipment, and data transmission modules at designated locations on the unmanned vessel and underwater robot to ensure that the equipment is not interfered with by the carrier itself during operation; establish a stable connection between the unmanned vessel and underwater robot and the onshore data processing center through a wireless communication network to transmit equipment status information and collected raw data in real time.

[0097] In this embodiment, the underwater topographic data acquisition process is as follows:

[0098] Unmanned surface vessel (USV) data acquisition process: The USV is activated and navigates within the waterway according to a preset path. The multibeam echo sounder continuously emits sound beams at a set high frequency (e.g., 200Hz) to measure water depth data in different directions in real time; the sonar equipment works simultaneously to perform high-resolution scanning of the underwater terrain and acquire acoustic image information of the terrain; the satellite positioning system on the USV updates its position information every second, and the inertial navigation system monitors attitude changes in real time to ensure that the collected data is accurately correlated with the actual geographical location;

[0099] Underwater robot data acquisition: For areas difficult for unmanned vessels to reach, such as shallow waters and reef-strewn areas, underwater robots are deployed. These robots use thrusters to flexibly adjust their position and attitude, approaching the target area to collect data. Multibeam echo sounders and sonar equipment acquire more detailed terrain data at close range, and their positioning information is exchanged with onshore base stations via an underwater acoustic positioning system to ensure the accuracy of the data's location.

[0100] Real-time data transmission and backup: The collected raw data is transmitted to the onshore data processing center in real time via a wireless communication link. At the same time, the unmanned vessel and underwater robot are equipped with large-capacity storage devices and perform local backups of the data at regular intervals (e.g., every 5 minutes) to prevent data loss or interruption during transmission.

[0101] It should be noted that the above description is merely a technical solution of the invention and not a limitation. Although the invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the invention without departing from the scope of the invention, and all such modifications and substitutions should be covered within the scope of the claims of the invention.

Claims

1. A method for improving channel traffic capacity based on underwater maps, characterized in that: Includes the following steps: S1 constructs high-precision underwater maps by deploying unmanned vessels and underwater robots, using multibeam echo sounding technology to collect underwater topographic data, and using satellite positioning and inertial navigation systems to ensure the accuracy of data collection locations. The raw data is then processed to construct high-precision underwater maps. S2 establishes a dynamic waterway planning model, collects hydrological data of different seasons and times of the waterway as well as historical ship passage data, and uses big data analysis and machine learning algorithms to generate dynamic waterway usage plans based on real-time hydrological conditions and ship passage needs. The S3 integrated ship navigation and management system includes the development of ship navigation terminal equipment, which integrates underwater map display, dynamic waterway planning information reception, and ship position positioning functions; it also includes the integration of dynamic waterway planning models with the monitoring system of waterway management departments to realize data interaction and collaborative management between ships and management departments; Step S2 includes the following steps: S21 multi-source data acquisition and integration works closely with hydrological monitoring departments to obtain detailed hydrological data of the waterway at different seasons and times. At the same time, it extracts historical vessel traffic data from the database of the waterway management department, classifies vessel size types, and determines the navigation space required by vessels in the waterway based on size parameters. A time synchronization mechanism is established to link hydrological data and vessel traffic data within the same time interval to form a complete data sample set D, providing a rich and accurate data foundation for subsequent model training. The model building and training of S22 machine learning uses the Long Short-Term Memory (LSTM) network algorithm in deep learning to build a dynamic waterway planning model. The data sample set D is input into the model for training, and the output is a recommended navigation route for different ship types under the current hydrological conditions. The algorithm is optimized to continuously adjust the model's weights and biases, so that the model can accurately predict reasonable navigation routes based on the input hydrological and ship information. The S23 model is updated and optimized in real time. At fixed time intervals, the latest hydrological data and vessel traffic data are obtained from hydrological monitoring departments and waterway management departments, and the new data is added to the data sample set D to form an expanded dataset. Incremental training is performed on the already trained LSTM model. During incremental training, the model structure remains unchanged, and only the model weights and biases are fine-tuned. At the same time, model performance evaluation metrics are introduced, and the training parameters of the model are dynamically adjusted according to the evaluation results to ensure that the model always maintains the best performance state and provides accurate and reliable support for real-time channel planning.

2. The method for improving channel traffic capacity based on underwater maps according to claim 1, characterized in that: Step S1 includes the following steps: S11 data acquisition involves deploying unmanned surface vessels (USVs) and underwater robots, equipped with multibeam echo sounders and sonar equipment. USVs efficiently conduct large-scale data acquisition in wide waterways, while underwater robots excel at penetrating narrow, complex, or shallow terrain areas for detailed exploration. The S12 data processing and map building process first denoises the collected data, retaining the effective low-frequency components, and then filters it. The filtering parameters are adjusted in real-time according to dynamic data changes to improve data stability and accuracy. Next, a feature-matching-based stitching algorithm is used to stitch the data together, extracting feature points from overlapping areas of data from different acquisition devices, constructing a feature matching matrix, aligning feature points from different datasets for seamless stitching, and building an underwater terrain model. Finally, the underwater terrain model is converted into a vector or raster format underwater map supported by a Geographic Information System (GIS), stored in a database, and a spatial index-based data retrieval mechanism is established to enable fast and efficient map data query and retrieval, meeting the application needs of real-time waterway management and ship navigation.

3. The method for improving channel traffic capacity based on underwater maps according to claim 1, characterized in that: Step S3 includes the following steps: The development of the S31 ship navigation terminal equipment is aimed at the complex and harsh navigation environment of ships. The core hardware processor of the terminal equipment is selected with a chip with high performance computing power and low power consumption. The screen adopts a high brightness and high contrast display. The positioning module integrates the Global Navigation Satellite System (GNSS) and Ultra-Wideband (UWB) positioning technology. The communication module supports multiple communication protocols. A navigation software system with multiple modules working together has been developed. The S32 waterway management system is integrated with the model, and an information receiving module is set up to parse the navigation instructions issued by the waterway management system. The instructions contain a series of coordinate points for the recommended route. i=1,2,L,n, by comparing with the ship's current position coordinates In contrast, the path planning algorithm is used to calculate the optimal navigation path from the ship's current position to the recommended route. The positioning module obtains the ship's position in real time and continuously tracks the ship's position by combining the trajectory estimation algorithm. Through continuous iteration, high-precision trajectory estimation is achieved, which can make up for the position monitoring when the positioning signal is briefly lost. The S33 system operates and is maintained through real-time two-way data interaction between the ship navigation terminal and the waterway management system. The waterway management system generates navigation instructions based on a dynamic waterway planning model and sends them to the ship navigation terminal via the communication network. Upon receiving the instructions, the ship displays them on its navigation software interface and plans its navigation path accordingly. Simultaneously, the ship uploads its own navigation status data to the waterway management system in real time. The waterway management system monitors ship navigation based on this data, performs global optimization management of waterway traffic, and establishes a system fault diagnosis and fault tolerance mechanism. When a fault occurs, fault tree analysis (FTA) technology is used to quickly locate the fault point. For some non-critical faults, the system employs fault-tolerant technology to ensure that the system can maintain basic operation even under certain fault conditions, guaranteeing the safety and continuity of waterway passage.

4. The method for improving channel traffic capacity based on underwater maps according to claim 2, characterized in that: In step S11, the unmanned vessel and underwater robot plan their navigation path based on the shape, width, and depth of the waterway and known terrain features. The multibeam echo sounder and sonar equipment work together to collect data, and the collected data can be transmitted in real time and backed up locally.

5. A method for improving channel traffic capacity based on underwater maps according to claim 2, characterized in that: In step S11, the multibeam echo sounder is based on acoustic wave reflection ranging. The emitted acoustic beam forms a certain opening angle θ with the horizontal plane. Assuming the speed of sound in water is v, it is corrected using an empirical formula: Where T is water temperature in °C; S is salinity; and D is depth in meters. After a multibeam echo sounder emits sound waves, it receives the signals reflected back from the underwater topography and measures the round-trip propagation time t of the sound waves. The water depth h at the corresponding measuring point can then be calculated using the formula... The calculations show that, simultaneously, the sonar equipment works in sync, transmitting broadband acoustic signals and receiving echo signals to construct an acoustic image of the underwater terrain.

6. The method for improving channel traffic capacity based on underwater maps according to claim 2, characterized in that: In step S11, to ensure high-precision positioning of the collected data, a satellite positioning system and an inertial navigation system are integrated. The satellite positioning system uses signals emitted by multiple satellites to determine the device position through trilateration. The inertial navigation system measures the acceleration and angular velocity of the device based on accelerometers and gyroscopes, and realizes coordinate transformation through attitude matrix to track the device's position and attitude changes in space in real time.

7. A method for improving channel traffic capacity based on underwater maps according to claim 3, characterized in that: In step S31, the hardware of the ship navigation terminal equipment is waterproof, dustproof, and shockproof, while the software enables underwater map display, dynamic waterway planning information reception, and ship position positioning.

8. The method for improving channel traffic capacity based on underwater maps according to claim 3, characterized in that: Step S32 also requires the establishment of an efficient communication network. Communication base stations are set up on shore, and a hybrid communication mode combining cellular and satellite communication is adopted. In areas with good signal coverage along the waterway, ships communicate with base stations via 4G / 5G cellular networks. In remote areas or areas with severe signal obstruction, the system switches to satellite communication to ensure the continuity of data transmission. Furthermore, ad-hoc communication is established between ships. When multiple ships are within close range, an ad-hoc network is automatically formed to exchange information and achieve efficient data forwarding between ships, ensuring timely information transmission between ships and improving the collaborative navigation capability of ships within the waterway.

9. A method for improving channel traffic capacity based on underwater maps according to claim 1, characterized in that: Step S22 includes the following steps: The S221 model receives the integrated data sample set D at the input layer, including water level H, water flow velocity V, water flow direction α, and ship type S. type Ship length L ship Ship width W ship Inside the LSTM network, the input feature vector X(t) at each time step t is compared with the hidden state at the previous time step. and cell state interaction; The S222 model LSTM unit passes through the input gate. Forgotten Gate and output gate To control the flow of information, the input gate calculation formula is: The formula for calculating the forget gate is: The formula for calculating the output gate is: Through the regulation of these gates, cell state Updated to New hidden state ; Where σ is the sigmoid activation function, This is the weight matrix. Let e ​​be the bias vector, and e denotes element-wise multiplication. The output layer of the S223 model, based on the hidden states processed by the LSTM network, outputs recommended navigation routes for different ship types under the current hydrological conditions. The navigation routes are represented by a series of discrete coordinate points, and a regression algorithm maps the hidden states to coordinate values. Let the coordinate points of the output navigation routes be... The model is trained by minimizing the mean squared error loss function between the predicted route and the historical best route. The model's weights and biases are continuously adjusted using optimization algorithms such as stochastic gradient descent, enabling the model to accurately predict reasonable navigation routes based on the input hydrological and ship information.

Images