Method and device for dredging congestion of inland waterway

By collecting multi-source data for spatiotemporal registration and fusion processing, and combining it with traffic flow prediction models to generate preventive control parameters, the problem of not being able to guide ships in advance in existing technologies has been solved, realizing intelligent dredging of inland waterways and improving traffic efficiency and stability.

CN121838525APending Publication Date: 2026-04-10CHINA COMM CONSTR FIRST HARBOR CONSULTANTS
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202610251789.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-03
Publication Date
2026-04-10

AI Technical Summary

Technical Problem

Existing technology cannot guide ships in advance, leading to increased congestion and low traffic efficiency in inland waterways.

Method used

By collecting multi-source traffic data and waterway meteorological and hydrological data, performing spatiotemporal registration and fusion processing, comprehensive traffic situation information is generated. Then, a trained traffic flow prediction model is used for forward-looking prediction to generate preventive control parameters, enabling individualized guidance for traffic lights and vessels.

Benefits of technology

It enables forward-looking prediction of waterway traffic conditions, avoids exacerbating congestion, and improves the overall traffic efficiency and operational stability of the waterway.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121838525A_ABST
    Figure CN121838525A_ABST
Patent Text Reader

Abstract

The invention discloses an inland waterway congestion dredging method and device, and relates to the field of traffic control, and the method comprises the steps: obtaining multi-source traffic data and waterway meteorological and hydrological data of each ship in a waterway; performing space-time registration and fusion processing on the multi-source traffic data, generating comprehensive traffic situation information of the whole section of the channel, and extracting traffic flow characteristic parameters and channel congestion indexes; inputting the traffic flow characteristic parameters, the channel congestion index and the channel meteorological and hydrological data into a trained traffic flow prediction model to obtain traffic prediction information of key nodes in the channel in a future preset time window; based on the comprehensive traffic situation information and the traffic prediction information, generating preventive control parameters for channel key node signal lamps; and according to the traffic prediction information and the preventive control parameters, traffic guidance information is generated and sent to the corresponding ship terminals, and the preventive control parameters are sent to the corresponding signal lamp control machines. The overall passing efficiency of the channel can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of traffic control technology, and in particular to a method and apparatus for alleviating congestion in inland waterways. Background Technology

[0002] In the rapid development of the shipping industry, inland waterways, as an important component of the comprehensive transportation system, have played an irreplaceable role in bulk cargo transportation, regional economic linkages, and ecological environmental protection due to their advantages such as large carrying capacity, low cost, low energy consumption, and low pollution. However, with the continuous growth of waterway transport volume and the increasing tonnage of vessels, inland waterways generally face natural limitations such as river bends and seasonal water level fluctuations. The navigation pressure at key nodes such as some pivotal sections, sluice gate areas, and narrow channels is becoming increasingly prominent, leading to vessel congestion and low traffic efficiency. Therefore, research on adaptive vessel guidance in waterways is particularly important to ensure safe navigation and regulate traffic order.

[0003] Currently, related technologies use radar and other detectors to monitor the queue length of ships in each direction at the junction, and dynamically adjust the green light duration of traffic lights based on the queue length to achieve channel signal control. However, this approach only makes post-congestion adjustments, which is a passive response. Its decision-making dimension is relatively singular and one-sided, and it cannot guide ships in advance, which can easily exacerbate congestion and lead to low ship passage efficiency. Summary of the Invention

[0004] The purpose of this application is to provide a method and apparatus for alleviating congestion in inland waterways, in order to solve the technical problem in the prior art where the inability to guide ships in advance exacerbates congestion and leads to low ship passage efficiency.

[0005] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for alleviating congestion in inland waterways, including: Acquire multi-source traffic data and meteorological and hydrological data of various vessels in the waterway; The multi-source traffic data is spatiotemporally registered and fused to generate comprehensive traffic situation information for the entire waterway. Based on this comprehensive traffic situation information, traffic flow characteristic parameters and a waterway congestion index are extracted. The comprehensive traffic situation information is used to characterize the distribution location, motion state, identity attributes, and overall waterway traffic status of all vessels in the waterway under a unified coordinate system. The waterway congestion index is used to reflect the overall smoothness of waterway traffic. The traffic flow characteristic parameters, waterway congestion index, and waterway meteorological and hydrological data are input into the trained traffic flow prediction model to obtain traffic prediction information for key nodes in the waterway within a future preset time window. Based on the comprehensive traffic situation information and the traffic prediction information, an adaptive signal control algorithm is used to generate preventive control parameters for signal lights at key nodes in the waterway. Based on the traffic forecast information and the preventive control parameters, traffic guidance information is generated for each vessel, and the traffic guidance information is sent to the corresponding vessel terminal. The preventive control parameters are also sent to the corresponding signal light controller.

[0006] Secondly, this application provides an inland waterway congestion relief device, which includes: The acquisition module is used to acquire multi-source traffic data and meteorological and hydrological data of various vessels in the waterway; The processing module is used to perform spatiotemporal registration and fusion processing on the multi-source traffic data to generate comprehensive traffic situation information for the entire waterway. Based on the comprehensive traffic situation information, traffic flow characteristic parameters and waterway congestion index are extracted. The comprehensive traffic situation information is used to characterize the distribution location, motion state, identity attributes, and overall waterway traffic status of all vessels in the waterway under a unified coordinate system. The waterway congestion index is used to reflect the overall smoothness of waterway traffic. The prediction module is used to input the traffic flow characteristic parameters, waterway congestion index and waterway meteorological and hydrological data into the trained traffic flow prediction model to obtain traffic prediction information of key nodes in the waterway in the future preset time window. The generation module is used to generate preventive control parameters for signal lights at key nodes of the waterway based on the comprehensive traffic situation information and the traffic prediction information using an adaptive signal control algorithm. The sending module is used to generate traffic guidance information for each vessel based on the traffic forecast information and the preventive control parameters, send the traffic guidance information to the corresponding vessel terminal, and send the preventive control parameters to the corresponding signal light controller.

[0007] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides a method and apparatus for alleviating congestion in inland waterways. Compared with existing technologies, this solution collects multi-source traffic data from vessels and meteorological and hydrological data from the waterway, and generates comprehensive traffic situation information through spatiotemporal registration and fusion processing. This achieves comprehensive perception of vessel distribution, movement status, identity attributes, and overall traffic status in the waterway, breaking through the limitations of related technologies that rely solely on radar monitoring of queue lengths. This provides comprehensive and accurate baseline data for waterway operation for subsequent decision-making. Furthermore, by inputting traffic flow characteristic parameters, waterway congestion index, and meteorological and hydrological data into a trained prediction model, it can obtain the traffic status at key nodes within a preset time window in advance, possessing the ability to predict waterway traffic conditions proactively. This upgrades signal control from a reactive, post-event response to a proactive, pre-event response. Preventative control measures are implemented to avoid exacerbating congestion. Based on comprehensive traffic conditions and forward-looking traffic forecasts, adaptive signal control algorithms are used to generate preventative control parameters for traffic lights. This goes beyond the reactive approach of adjusting green light durations after congestion occurs; instead, signal parameters are optimized in advance by anticipating potential congestion, preventing congestion from escalating at its source. Furthermore, individualized traffic guidance information is generated for each vessel and pushed to the terminal by combining traffic forecasts and preventative control parameters. This achieves two-way coordination between traffic lights and proactive vessel guidance, enabling not only precise and intelligent control of traffic lights but also advance path and speed guidance for vessels. This solves the problems of incomplete technical decision-making and the inability to provide advance guidance, which lead to increased congestion and low traffic efficiency. As a result, the overall traffic efficiency and operational stability of the waterway are significantly improved. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A schematic diagram of the structural environment for the application environment of the inland waterway congestion relief method provided in an embodiment of this application; Figure 2 A schematic diagram of the structure of an inland waterway congestion relief system provided in another embodiment of this application; Figure 3 A flowchart illustrating an embodiment of the inland waterway congestion relief method provided in this application; Figure 4 A flowchart illustrating a method for constructing a digital waterway traffic panorama, provided as another embodiment of this application; Figure 5 A schematic diagram illustrating the process of prediction using a traffic flow prediction model, provided as an embodiment of this application; Figure 6A flowchart illustrating the training of a traffic flow prediction model provided in one embodiment of this application; Figure 7 A schematic diagram illustrating the process of prediction using a traffic flow prediction model, provided for another embodiment of this application; Figure 8 A flowchart illustrating an embodiment of the adaptive signal control method based on reinforcement learning provided in this application; Figure 9 A schematic diagram of the functional modules of an inland waterway congestion relief device provided in an embodiment of this application; Figure 10 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0010] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] For ease of understanding, some technical terms involved in the embodiments of this application are explained below: An Automatic Identification System (AIS) is a maritime navigation and communication device installed on ships. Utilizing the Global Positioning System (GPS) and communication technology, it enables ships to automatically transmit information such as their position, course, and speed, and exchange this information in real time with surrounding vessels and shore-based stations. Used for collision avoidance, it automatically and continuously broadcasts static, dynamic, voyage, and safety-related information to surrounding vessels and shore-based stations via the Very High Frequency (VHF) radio band. It is an important navigational aid and monitoring tool in the maritime field.

[0013] Green wave traffic: Green wave traffic, a traffic control technology that utilizes traffic engineering, coordinates the time difference of traffic lights at adjacent intersections on main roads to allow vehicles to pass through green lights continuously while traveling at a set speed.

[0014] Ultra-Reliable and Low-Latency Communications (uRLLC) is designed to meet the demands for low latency and high reliability in applications such as real-time control and autonomous driving. The main characteristics of uRLLC are extremely low latency and high reliability, ensuring the reliability and stability of real-time and mission-critical applications.

[0015] Related technologies use radar and other detectors to monitor the queue length of ships in each direction at the junction, and dynamically adjust the green light duration of the signal lights based on the queue length to achieve channel signal control. However, this approach only makes post-hoc adjustments to existing congestion, which is a passive response. Furthermore, its decision-making dimension is relatively singular and one-sided, failing to consider the spatiotemporal evolution of traffic flow throughout the channel, and it cannot provide advance guidance for ships. Lacking predictability, it is highly likely to exacerbate congestion and result in low ship passage efficiency.

[0016] To address the aforementioned shortcomings, this application provides a method for alleviating congestion in inland waterways. Compared to existing technologies, this solution collects multi-source vessel traffic data and waterway meteorological and hydrological data, and generates comprehensive traffic situation information through spatiotemporal registration and fusion processing. This achieves a comprehensive perception of vessel distribution, movement status, identity attributes, and overall traffic conditions within the waterway, overcoming the limitations of related technologies that rely solely on radar monitoring of queue lengths. This provides comprehensive and accurate baseline data for subsequent decision-making regarding waterway operations. Furthermore, by inputting traffic flow characteristic parameters, waterway congestion index, and meteorological and hydrological data into a trained prediction model, it can obtain traffic conditions at key nodes within a preset time window in advance, possessing the ability to predict waterway traffic conditions proactively. This upgrades signal control from a reactive, reactive approach to a proactive, preventative control, avoiding further congestion. Based on comprehensive traffic conditions and forward-looking traffic forecast information, an adaptive signal control algorithm is used to generate preventive control parameters for traffic lights. This is no longer limited to the delayed operation of "adjusting green light duration after congestion," but rather optimizes signal parameters in advance by predicting potential congestion, thus preventing congestion from escalating at the source. Furthermore, by combining traffic forecast information and preventive control parameters, individualized traffic guidance information is generated for each vessel and pushed to the terminal, achieving two-way coordination between traffic lights and proactive vessel guidance. This not only enables precise and intelligent control of traffic lights but also provides advance path and speed guidance for vessels. It solves the problems of congestion aggravation and low traffic efficiency caused by one-sided technical decision-making and the inability to provide advance guidance, significantly improving the overall traffic efficiency and operational stability of the waterway.

[0017] This application provides a method for alleviating congestion in inland waterways, which can be applied to, for example... Figure 1The application environment of the inland waterway congestion mitigation method is shown. This application environment includes an inland waterway congestion mitigation system, which comprises: a congestion mitigation device 10, at least one vessel terminal 20, and at least one signal light controller 30. The congestion mitigation device 10 establishes communication connections with each vessel terminal 20 and each signal light controller 30. The at least one vessel terminal may include a first vessel terminal, a second vessel terminal, ..., an nth vessel terminal, and the at least one signal light controller may include a first signal light controller, a second signal light controller, ..., an mth signal light controller, where n and m are both positive integers.

[0018] The congestion mitigation device 10 acquires multi-source traffic data and meteorological and hydrological data from various vessels in the waterway, generates traffic guidance information for each vessel, and sends the traffic guidance information to the corresponding vessel terminal. It also generates preventative control parameters and sends them to the corresponding signal light controller. The vessel terminal 20 receives the traffic guidance information and navigates accordingly. The signal light controller 30 receives the preventative control parameters and controls the signal lights according to those parameters.

[0019] Optionally, please see Figure 2 As shown, the congestion mitigation device 10 includes a sensing module 11, a communication transmission module 12, and a data processing and control center 13. The traffic light controller 30 includes a signal receiving module 31 and a signal execution module 32. The sensing module 11 may include various data acquisition devices, such as millimeter-wave radar, AIS base stations, and cameras. The communication transmission module 12 establishes communication connections with the sensing module 11 and the data processing and control center 13, respectively.

[0020] The sensing module 11 is used to collect multi-source traffic data from various vessels in the waterway; the communication transmission module 12 is used to realize data interaction between various modules in the system. The data processing and control center may include a traffic situation fusion unit, a short-term traffic flow prediction unit, an adaptive signal control unit, and a macro-guidance decision unit. The traffic situation fusion unit is used to generate comprehensive traffic situation information based on the acquired multi-source traffic data; the short-term traffic flow prediction unit is used to load and run a short-term traffic flow prediction model to perform traffic prediction; the adaptive signal control unit is used to run an adaptive signal control algorithm to generate preventive control parameters for signal lights at key nodes in the waterway; and the macro-guidance decision unit is used to generate speed guidance information and path recommendation information.

[0021] The short-term traffic flow prediction unit and the adaptive signal control unit can be deployed on a server or terminal to ensure the real-time performance of model inference and control decisions. The path recommendation information generated by the macro-guidance decision unit is sent to the corresponding ship terminal through the communication transmission module and displayed graphically on the electronic river map interface of the ship terminal.

[0022] The signal execution module inside the signal light controller 30 can be set at key nodes in the waterway, such as junctions, to execute signal light control commands corresponding to preventative control parameters. The ship terminal 20 is installed on the ship to receive and display speed guidance information and route recommendation information.

[0023] The system provided in this embodiment has a high degree of integration and maximizes information utilization. By integrating functions such as perception, prediction, control, and guidance, it breaks down information silos. Through deep fusion and mining of multi-source data, it maximizes the value of data and builds a complete, unified, and efficient smart waterway management solution system, reducing the complexity of operation and maintenance and providing a powerful decision support tool for ship management.

[0024] The aforementioned terminal can be a terminal device in various AI application scenarios. For example, the terminal can be a smart home device such as a smart TV or smart TV set-top box; or it can be a mobile portable terminal such as a smartphone, tablet, or e-book reader; or it can be a smart wearable device such as smart glasses or a smartwatch. This embodiment does not specifically limit this. The server can be a single server, a server cluster consisting of several servers, or it can contain one or more virtualization platforms, or it can be a cloud computing service center. The server can be a server device that provides backend services for the AI ​​applications installed on the aforementioned terminal.

[0025] A communication connection is established between the terminal and the server via a wired or wireless network. Optionally, the aforementioned wireless or wired network uses standard communication technologies and / or protocols. The network is typically the Internet, but can also be any network, including but not limited to a Local Area Network (LAN), Metropolitan Area Network (MAN), Wide Area Network (WAN), mobile, wired or wireless network, private network, or any combination of virtual private network.

[0026] Optionally, the system may also include a cloud platform that can be connected to congestion mitigation devices, ship terminals, and signal light controllers. The cloud platform is used to store historical traffic data, various model parameters, and system logs, and provides a human-machine interface for remote monitoring and management.

[0027] In one exemplary embodiment, such as Figure 3As shown, a method for alleviating congestion in inland waterways is provided. This method is executed by a congestion mitigation device, which can be executed independently by a computer device such as a terminal or server, or jointly by a terminal and a server. In this embodiment, the method is described using a server as an example, and includes the following steps S201 to S205. Wherein: Step S201: Obtain multi-source traffic data and meteorological and hydrological data of each vessel in the waterway.

[0028] It should be noted that the aforementioned multi-source traffic data refers to various types of data acquired through multiple data acquisition devices that reflect the operational status of waterway traffic, including: Automatic Identification System (AIS) information, vessel visual information, and target detection information. Multi-source traffic data can be acquired through multiple data acquisition devices. To achieve accuracy and comprehensiveness in target perception within the waterway, data acquisition devices are deployed at key nodes such as waterway junctions, narrow river sections, and dam approach channels. These multiple data acquisition devices can include, for example, radar, AIS base stations, and cameras. Vessel visual information is obtained by cameras capturing images of vessels in the waterway and can include hull outlines and dimensions. Target detection information is obtained by radar, and vessel attribute information can be obtained by AIS base stations. The radar can be millimeter-wave radar. Vessel attribute information can include: the position, speed, heading, and identification of each vessel in the waterway.

[0029] Understandably, AIS information is transmitted via Very High Frequency (VHF) radio waves and encapsulated using the NMEA 0183 protocol format. Common AIS message types include static information (such as ship name and MMSI number), dynamic information (such as position and speed), and safety-related information (such as turning intention). The MMSI number is a unique identifier for a ship in the AIS system, assigned by the International Telecommunication Union (ITU) and cannot be arbitrarily changed.

[0030] Optionally, the multi-source traffic data and waterway meteorological and hydrological data of various vessels in the aforementioned waterway can also be obtained from external devices or imported from blockchain or databases.

[0031] As one feasible approach, a unified waterway geographic coordinate system can be established. Taking millimeter-wave radar as an example, the aforementioned target detection information includes heading information, the ship's first position information, and velocity vector. The heading information can include the heading angle. The millimeter-wave radar scans a fan-shaped area of ​​the waterway at a preset scanning frequency to obtain point cloud data of all ship targets within the fan-shaped area. Based on the point cloud data, target clustering and tracking algorithms are used to analyze and process the point cloud data, obtaining the first position information of each ship within the waterway coverage area. ), velocity vector and heading angle The first location information is based on the waterway geographic coordinate system.

[0032] The aforementioned AIS information includes: the vessel's second position information (MMSI), ground speed, identification information, and static attribute information. Static attribute information may include vessel type, tonnage, etc. The identification information and related data of each vessel are collected by the Automatic Identification System (AIS) base station. Specifically, this can be achieved by the data processing and control center receiving periodically broadcast messages from each vessel's AIS terminal and parsing these messages to obtain the vessel's identification information (MMSI), second position information (GSI), etc. Ground speed Ground heading Ship type and ship tonnage Data such as...

[0033] During the data acquisition process using cameras, which can be high-definition cameras and deployed at key nodes in the waterway, real-time video streaming of these key nodes is captured. Keyframes are then identified from the video stream, and image recognition algorithms are used to process these keyframes to obtain the ship's visual information. This camera-based acquisition of ship visual information enables visual verification, aiding in the authentication of targets detected by millimeter-wave radar and AIS, eliminating false targets. It also allows for behavior recognition, identifying abnormal ship behaviors such as yaw rate and illegal anchoring. Furthermore, it enables status recording to ensure the acquisition of ship visual characteristics even in low-visibility conditions.

[0034] After acquiring ship visual information, AIS information, and target detection information, the ship visual information captured by the camera can be cross-validated with the target detection information detected by millimeter-wave radar and the AIS parameters reported by AIS. By calculating the positional deviation and size matching degree of the visual targets collected by the camera, the target detection information collected by the radar, and the AIS targets, false targets caused by radar clutter and false AIS signal alarms are eliminated to ensure the authenticity and reliability of target identification. False targets may include, for example, floating objects and signal interference points.

[0035] Furthermore, the aforementioned waterway meteorological and hydrological data refers to real-time dynamic and historical statistical data reflecting the weather conditions and aquatic environmental characteristics of the waterway area. These are core environmental parameters for inland waterway navigation environment assessment and signal guidance strategy optimization. Weather conditions in the waterway area include rainfall, visibility, and wind speed, while aquatic environmental characteristics include water level, flow velocity, and water depth. For example, wind speed can be collected in real time through meteorological stations and hydrological sensors deployed along the waterway. ,wind direction ,visibility and water flow speed Data. Among them, meteorological stations and hydrological sensors may include: anemometers, laser visibility meters, Doppler current meters, and pressure level gauges.

[0036] In this embodiment, by acquiring multi-source traffic data from various vessels in the waterway, more comprehensive multi-dimensional information about the vessels can be collected, providing good data guidance for subsequent congestion management. Furthermore, by acquiring waterway meteorological and hydrological data, it is easier to subsequently revise the vessel motion model and traffic flow prediction model based on this data, making the management strategy closer to the actual navigation environment.

[0037] Step S202 involves performing spatiotemporal registration and fusion processing on multi-source traffic data to generate comprehensive traffic situation information for the entire waterway. Based on this comprehensive traffic situation information, traffic flow characteristic parameters and waterway congestion index are extracted.

[0038] Comprehensive traffic situation information is used to characterize the distribution, movement, identity attributes, and overall traffic status of all ships in the waterway under a unified coordinate system; the waterway congestion index is used to reflect the overall smoothness of waterway traffic.

[0039] Understandably, in order to ensure the data quality of AIS information and avoid abnormal data from interfering with subsequent fusion processing and decision analysis, it is necessary to preprocess the AIS information acquired by AIS base stations to remove abnormal data points that clearly exceed physical limits. This preprocessing may include rationality verification and filtering, and parameter standardization mapping.

[0040] During the process of rationality verification and filtering, physical constraint verification and sliding window filtering algorithms can be used to eliminate abnormal data points caused by signal packet loss and equipment failure; and linear interpolation can be used to fill in missing data to ensure data continuity; abnormal data points include instantaneous speed changes and position jumps.

[0041] During the parameter standardization mapping process, the ship type in the AIS information can be... With tonnage Information is mapped to standardized classification labels and numerical ranges required for subsequent traffic flow analysis and prediction models. Vessel types include passenger ships, cargo ships, dangerous goods ships, and emergency ships. In generating standardized classification labels, this can be done as "passenger ship = 1, cargo ship = 2, dangerous goods ship = 3, emergency ship = 4," facilitating subsequent priority ranking. Vessel tonnage is divided into multiple numerical ranges according to inland waterway shipping standards, such as ≤500 tons, 501-1000 tons, 1001-5000 tons, and >5000 tons, converting them into quantitative parameters that the model can directly use, improving data adaptability.

[0042] In one embodiment, a specific implementation method is also provided for spatiotemporal registration and fusion processing of multi-source traffic data to generate comprehensive traffic situation information for the waterway. Please refer to [link to relevant documentation]. Figure 4 As shown, the method includes: Step S301: Establish a waterway geographic coordinate system and convert the first position information, the second position information, and the field of view range information of the camera into the waterway geographic coordinate system.

[0043] Step S302: Establish a unified timestamp for AIS information, ship visual information, and target detection information, and use the timestamp as a reference to perform time synchronization and alignment processing on asynchronous data in multi-source traffic data for time registration.

[0044] Step S303: After time registration is completed, the first position information and velocity vector are used as observations, and the second position information and ground speed are used as auxiliary correction quantities. A data fusion algorithm is used to perform fusion filtering to obtain the fused trajectory data of each ship. The fused trajectory data includes: the optimal estimated position and the optimal estimated speed.

[0045] Step S304 involves associating and binding the fused trajectory data with identity identifiers and static attribute information to assign dynamic and static attribute information to each vessel.

[0046] Step S305: After the attributes are assigned, a digital waterway traffic panorama is constructed in a unified coordinate system based on the fused trajectory data and dynamic and static attribute information of all ships.

[0047] Specifically, after acquiring ship visual information and target detection information, and preprocessing AIS information, spatial registration of multi-source traffic data is required. This involves obtaining the field of view information of the cameras, which can be the physical coverage area information calculated based on the camera's installation location, pitch angle, and focal length. In the spatial registration process, taking millimeter-wave radar and high-definition cameras as an example, a unified waterway geographic coordinate system is first constructed. This coordinate system could be, for example, the WGS-84 geodetic coordinate system or a local plane coordinate system. Then, through coordinate transformation algorithms such as affine transformation and projection mapping, the first location information acquired by the millimeter-wave radar (…) is converted into… ), the second location information reported by the AIS base station ( The information, including the field of view of the high-definition camera, is mapped to this unified coordinate system to eliminate spatial positioning deviations between different devices and achieve spatial alignment of the same target ship at the same coordinate.

[0048] After spatial registration is completed, a unified timestamp is established for AIS information, ship visual information, and target detection information, such as using UTC time or local standard time, so as to time-stamp the data streams of all sensing devices such as radar, AIS base stations, and cameras. For the data transmission delay and sampling frequency differences of different devices, such as radar sampling frequency of 10Hz and AIS sampling frequency of 1Hz, asynchronous data can be identified. A linear interpolation method or sliding time window method is used to time-synchronize and align asynchronous data to ensure that a panoramic data snapshot of the waterway containing data from all sensors is formed at the same timestamp, avoiding target association errors caused by time differences.

[0049] After time registration is completed, taking millimeter-wave radar as an example, a fusion algorithm based on Kalman filtering is adopted, combined with target association and identity binding, to achieve integrated integration of ship dynamic and static data. During the Kalman filtering data fusion process, due to the high real-time performance and high sampling rate of radar data, the first position information of the millimeter-wave radar (… ) and velocity vector As a core observation, due to the advantages of high stability and no obstruction interference in AIS data, the second location information in AIS information ( and ground speed As an auxiliary correction factor, the sensor measurement noise is dynamically corrected through the prediction-update iterative process of Kalman filtering, and the optimal estimated position of each ship is output. ) and optimal estimated speed This fusion result balances the real-time performance of radar with the accuracy of AIS. Compared to single-sensor data, it reduces position error by more than 30% and improves velocity stability by 50%, effectively avoiding measurement biases inherent in single-device systems. Among these, the velocity vector... This includes speed and heading.

[0050] In the process of fusing the first position information of the ship acquired by millimeter-wave radar and the second position information acquired by AIS base station, it can be determined whether the position deviation between the two is within a preset deviation threshold, and whether the ship's heading and speed parameters are consistent. If both conditions are met, the ship is identified as the same target, and a more accurate optimal estimated position is generated through data fusion algorithm. ) and optimal estimated speed This allows for the integration of the advantages of both types of equipment, such as the real-time performance of radar and the precise identity association of AIS, resulting in more accurate position and velocity information. The preset deviation threshold can be customized according to actual needs, for example, 50 meters. The data fusion algorithm can be, for example, a Kalman filter algorithm.

[0051] For targets whose positional deviation between the first and second positional information exceeds a preset deviation threshold, whose headings or speeds are inconsistent, or whose targets are detected by only a single device (such as small vessels without AIS signals that are only identified by radar, or vessels in radar blind spots that are only reported by AIS), a temporary identity identifier can be temporarily assigned, marked separately, and continuously tracked. The identity information can be supplemented later through camera visual recognition or VHF interaction. The separate markings, such as "radar target only" or "AIS target only," can be handled as special cases to ensure that all vessel targets in the waterway are effectively monitored, avoid target omissions due to data matching failures, and ensure the integrity of comprehensive traffic situation information.

[0052] The fused trajectory data (optimal estimated position and optimal estimated speed) of each vessel is linked one-to-one with the unique vessel identification code (MMSI code) and static attribute information parsed from AIS information, thereby assigning dynamic and static attribute information to each vessel. The static attribute information includes the vessel type. ,tonnage Length, width, etc.

[0053] In this step, based on the information collected from the ships, the ships can be processed according to different situations, thereby ensuring that each traffic entity has a complete data profile of dynamic trajectory and static attributes, providing data support for subsequent priority ranking and strategy optimization.

[0054] After assigning dynamic and static attributes, a visualized and quantifiable digital waterway traffic panorama is constructed in a unified coordinate system based on the optimal estimated position, optimal estimated speed, and bound dynamic and static attribute information of all vessels. This digital waterway traffic panorama can present the distribution, movement, and identity types of vessels throughout the entire waterway in real time. The panorama supports dynamic updates and can intuitively reflect the real-time situation of vessels entering and leaving the monitoring area, speed changes, and encounters, providing operators with a global perspective of navigation status monitoring and a data carrier for subsequent feature extraction. The vessel distribution includes the number and density distribution of vessels in each river section, and the movement status includes each vessel's heading, speed, and expected navigation trajectory.

[0055] In one embodiment, based on comprehensive traffic situation information, key traffic flow characteristic parameters and waterway congestion index are extracted, including: Within a preset time interval, the number of vessels at key sections of the waterway is counted, and the traffic flow per unit time is calculated based on the number of vessels. The optimal estimated speeds of all vessels within the key area of ​​the waterway are arithmetically averaged to obtain the average speed of the key area. At a preset time, the sum of the projected lengths of all vessels within the waterway segment is obtained and compared with the total length of the waterway segment to obtain the waterway occupancy rate. The total number of vessels and the area of ​​the key area are obtained, and the vessel density is calculated based on the total number of vessels and the area. The traffic flow, average speed, waterway occupancy rate, and vessel density are weighted and fused to obtain the waterway congestion index.

[0056] It should be noted that the aforementioned preset time intervals can be dynamically adjusted based on actual channel traffic flow, such as 5 minutes or 10 minutes. Key channel sections include upstream sections of confluences and dam entrance sections, and key channel areas include the confluence impact zone and narrow river sections. The aforementioned traffic flow characteristic parameters include: traffic flow, average speed, channel occupancy, and vessel density. Traffic flow reflects the level of traffic congestion in the channel; average speed reflects traffic efficiency; channel occupancy directly reflects the utilization level and congestion status of channel space, for example, a channel occupancy rate > 70% can be considered severe congestion; vessel density quantifies the density of vessels in the channel space, assisting in assessing congestion risk.

[0057] Specifically, after obtaining a digital panoramic view of waterway traffic, four key traffic flow characteristic parameters are extracted from the digital panoramic view to achieve a quantitative description of navigation status. Based on a set fixed time interval Δt, the total number N of vessels passing through key sections of the waterway is counted, and then calculated using the formula... Calculate traffic flow per unit time (e.g., ships / hour). This is used when calculating average speed. During the process, key areas of the waterway are selected, such as the influence zone of confluences and narrow river sections. The optimal estimated speed of all vessels within these key areas is obtained, and the arithmetic mean of all optimal estimated speeds is calculated to obtain the average speed of the area, expressed in km / h. Lower speeds indicate lower traffic efficiency and a potential for congestion. This process is then used to determine the waterway occupancy rate. During the process, at a preset specific time, the sum of the projected lengths of all ships within a key channel segment and the total length of that channel segment can be obtained. The sum of the projected lengths represents the projection of the ship lengths along the channel axis, and the ratio between the two can be calculated and expressed as a percentage. This ratio is then used to calculate ship density. During the process, the total number of ships and the area of ​​the key region are obtained, the average number of ships per unit area in the key waterway region is calculated, and the formula is used to obtain the total number of ships. The calculation is performed, where M is the total number of ships within the critical area, and Area is the area of ​​that critical area, in units of ships / km². 2 .

[0058] After extracting all traffic flow characteristic parameters, in order to comprehensively assess the overall smoothness of waterway traffic, a weighted fusion of the above four types of characteristic parameters is performed to calculate the comprehensive waterway congestion index. The weights of each parameter can be determined based on their influence on the results. For example, traffic flow has a weight of 0.2, average speed has a weight of 0.3, channel occupancy has a weight of 0.3, and ship density has a weight of 0.2. These weights can be dynamically adjusted based on channel characteristics. After standardization, a weighted sum is obtained to obtain the channel congestion index, which has a value range of [0,1]. This channel congestion index can reflect the smoothness of the channel. For example, CI < 0.3 indicates smooth flow, 0.3 ≤ CI < 0.6 indicates mild congestion, and CI ≥ 0.6 indicates severe congestion.

[0059] In this step, by normalizing and integrating multi-dimensional traffic flow characteristics, a waterway congestion index is obtained, which can quantitatively represent the overall smoothness of waterway traffic and provide an intuitive and quantitative decision-making basis for the adaptive adjustment of subsequent traffic management strategies.

[0060] Step S203: Input traffic flow characteristic parameters, waterway congestion index and waterway meteorological and hydrological data into the trained traffic flow prediction model to obtain traffic prediction information for key nodes in the waterway in the future preset time window.

[0061] It should be noted that the traffic flow prediction model described above is used to predict traffic conditions within a future preset time window. This model can be an LSTM model, and the preset time window can include K future time slices. Traffic prediction information includes: predicted traffic volume, predicted average air speed, and predicted congestion index.

[0062] After obtaining traffic flow characteristic parameters and waterway congestion index, these are input into a trained traffic flow prediction model to obtain traffic prediction information for key nodes in the waterway within a future preset time window.

[0063] In one embodiment, traffic flow characteristic parameters, waterway congestion index, and waterway meteorological and hydrological data are input into a trained traffic flow prediction model to obtain traffic prediction information for key nodes in the waterway within a future preset time window, including: Traffic flow characteristic parameters and waterway congestion index of key nodes in the waterway over the past T time slices are normalized and preprocessed to obtain processed data; T and K are both positive integers. The processed data and the corresponding waterway meteorological and hydrological data are used as input sequences. Through a trained traffic flow prediction model, the predicted parameters are output and inversely normalized to obtain the traffic flow, predicted average speed, and predicted congestion coefficient for the next K time slices. The traffic flow prediction model includes an input layer, at least one hidden layer, and a fully connected output layer connected in sequence. The input layer is used to receive the input sequence, the hidden layer is used to learn the long-term and short-term dependencies in the traffic flow characteristic parameters, and the fully connected output layer is used to map the final output of the hidden layer into traffic prediction information. The traffic prediction information is represented by a traffic prediction sequence.

[0064] In this embodiment, a time-series sample is constructed based on traffic flow feature parameters from the past T time slices. Represented by a sequence of traffic flow feature parameters, it is a multi-dimensional time-series tensor used to capture the temporal dependencies of traffic flow, such as peak-hour periodicity and ship traffic continuity. T is the length of the historical data window; for example, T=12 corresponds to a 5-minute time slice interval, covering data from the past hour. The time slice interval is consistent with the time granularity of the data feature extraction, ensuring data continuity. The future prediction time window includes K future time slices, where K is the length of the future prediction time window; for example, K=6 corresponds to the next 30 minutes. The value of K is consistent with the decision cycle of adaptive signal control. For example, if the signal is adjusted every 5 minutes, the prediction of the next 30 minutes can support 6 decisions, avoiding a disconnect between prediction and application.

[0065] Please see Figure 5 As shown, the traffic flow characteristic parameters of the past T time slices, i.e., the historical data [tT,t], constitute the input sequence. This input sequence is then used to perform time-series prediction through a trained LSTM model, thereby obtaining the prediction results for the next K time slices, i.e., the prediction results [t+1,t+K], which are represented by a traffic state prediction sequence. Specifically, in each historical time slice... The input feature vector should include at least: traffic flow for that historical time slice. Average speed Ship density and waterway congestion index The traffic flow characteristic parameters from the past T time slices are concatenated in chronological order to form a multidimensional time series tensor. As the core input of the model, .

[0066] Alternatively, external feature vectors can be introduced when constructing the model input sequence. This external feature vector includes waterway meteorological and hydrological data strongly correlated with traffic flow, to improve the model's adaptability to sudden environmental changes. Each historical time slice... Corresponding external feature vector E Including historical time films wind speed With visibility Wind speed affects ship speed and maneuverability, while visibility directly determines the channel's capacity. These two parameters are key environmental factors influencing sudden changes in inland waterway traffic flow; for example, dense fog can cause ships to slow down and traffic volume to drop sharply. By comprehensively considering the meteorological and hydrological data of this waterway, the model's predictive accuracy and robustness under complex and abrupt changes can be enhanced, making the predictions more realistic and providing a more reliable basis for subsequent intelligent control.

[0067] external feature vectors The input vector is formed by concatenating the multidimensional time series tensor of the corresponding time slice. This input sequence is then expanded into a "T×6" form, enabling the model to learn the coupling relationship between traffic flow and the environment, thereby enhancing the robustness of prediction.

[0068] The input sequence is processed by a trained traffic flow prediction model to obtain traffic prediction information for the next K time slots, which is represented by a traffic state prediction sequence. Each prediction time slot... The output vector includes predicted traffic flow. Predicted average speed and predicted congestion index This sequence constitutes the predicted result output by the model: ; .

[0069] The traffic flow prediction model described above can be an LSTM-based time-series prediction network used to capture long-term dependencies. It includes an input layer, at least one hidden layer, and a fully connected output layer. The input layer receives the constructed input sequence, enabling standardized data access. The LSTM hidden layer is the core feature learning module, with at least one layer; for complex scenarios, two to three layers can be stacked. Gating mechanisms (input gate, forget gate, output gate) effectively capture long-term dependencies in traffic flow, such as the correlation between the daily traffic peak at 8 AM and the previous day's traffic distribution. The number of neurons in the hidden layer can be optimized based on the data scale, for example, 128 or 256 neurons, to avoid overfitting or underfitting. The fully connected output layer maps the high-dimensional features output by the LSTM hidden layer to the predicted target shape, resulting in a traffic prediction sequence. The output dimension is (K,3), where K is the prediction window length and 3 is the number of output parameters per time slice. LSTM can solve the gradient vanishing problem in training long-sequence data and is more suitable for capturing the periodic changes and sudden fluctuations in traffic flow from peak to off-peak to trough.

[0070] The traffic flow prediction model described above can be constructed through the following steps: Historical waterway traffic flow characteristic data is acquired and normalized preprocessed to obtain processed historical data. The processed historical data is then divided into training and validation sets according to a preset ratio. The processed historical data includes real sequences. The processed historical data is input into an initial model for time series prediction, and the predicted sequence is output. Based on the predicted and real sequences, a loss function is constructed. By minimizing the loss function, an optimizer is used to iteratively update the parameters in the initial model to obtain the trained model. The validation set is then used to verify the performance of the trained model to construct a traffic flow prediction model.

[0071] Specifically, please see Figure 6 As shown, in the process of building the traffic flow prediction model, historical waterway traffic flow characteristic data are collected to construct training samples. The input of these training samples is: The corresponding output sequence is: S= ; When introducing external features, the external feature vector is concatenated into the input historical sequence. Because the input features have vastly different dimensions, direct training would lead to model weights biased towards features with large vectors. Therefore, normalization preprocessing of the training data is necessary to obtain processed historical data. This processed historical data is then divided into training and validation sets according to a preset ratio. The LSTM model is initialized and the model training loop begins. Forward propagation is used to calculate the output prediction sequence. The mean squared error between the predicted and true sequences is used as the loss function. The Adam optimizer is used for iterative training, minimizing the loss function and updating the model weights in reverse. The trained model is then obtained, and the validation set is input to monitor its performance. Based on the performance parameters, early stopping is determined (i.e., whether the validation set loss has significantly decreased). When the validation set loss no longer significantly decreases, an early stopping strategy is adopted to prevent overfitting, and the optimal model parameters are saved. Training then ends, resulting in the traffic flow prediction model. The normalization preprocessing can use the Min-Max method for normalization, applying normalization to each feature dimension. (like Handle independently. flow, Average speed, For ship density, The waterway congestion index can be expressed by the following formula: ; in, , These are the minimum and maximum values ​​of this feature on the training set.

[0072] After obtaining the processed historical data through preprocessing, when the processed historical data is used to perform time series prediction and output the predicted sequence through the initial model, the output predicted sequence can be inversely normalized through inverse operation to restore the original dimensions. The predicted value output by the model is the normalized value in the interval [0,1].

[0073] Please see Figure 7 As shown, after training the traffic flow prediction model, online prediction can be performed. Real-time data is received, including traffic flow characteristic parameters for the current and historical T time slices. This data is then preprocessed, undergoing the same normalization process as the training data, to obtain processed data. When external features are introduced, such as waterway meteorological and hydrological data, the processed data (E) is then stitched together with the real-time waterway meteorological and hydrological data. The input sequence is loaded into a pre-trained LSTM prediction model, and forward inference is performed to output the predicted sequence for the next K time slices. .

[0074] The predicted sequence is then denormalized to restore the data dimensions, yielding traffic flow, predicted average speed, and predicted congestion coefficient for the next K time slices.

[0075] This embodiment introduces a short-term traffic flow prediction model, using real-time traffic flow characteristics, congestion index, and meteorological and hydrological data as model inputs. This achieves a key shift in waterway signal control from "passive response" to "active prevention." It can accurately output traffic prediction information for key nodes within a preset future time window, breaking the limitations of traditional passive responses relying solely on current data. This allows the system to anticipate potential congestion trends in advance and implement targeted control measures (such as extending the green light window to ensure rapid vessel passage and guiding vessels to divert and avoid potential congestion points), rather than passively adjusting after congestion occurs. This not only provides forward-looking data for preventative signal control, avoiding the formation and exacerbation of congestion, but also provides a reference for future traffic conditions for vessel speed guidance and route recommendations, making guidance strategies more aligned with actual traffic needs. Ultimately, this improves waterway efficiency through both signal control and vessel behavior, solving problems such as decision-making lag and chaotic traffic order caused by the lack of predictive capabilities in related technologies. Furthermore, this proactive prevention mode effectively reduces the time lost by ships due to congestion and waiting, and is expected to reduce the average waiting time of ships at junctions by 15%-30%. At the same time, it optimizes the allocation of waterway resources from a global perspective, thereby increasing the overall capacity of the waterway by more than 20%. It completely solves the pain points of congestion aggravation and low traffic efficiency under the traditional passive response mode, and greatly improves the efficiency and stability of waterway operation.

[0076] Step S204: Based on comprehensive traffic situation information and traffic prediction information, an adaptive signal control algorithm is used to generate preventive control parameters for signal lights at key nodes of the waterway.

[0077] It should be noted that the aforementioned preventive control parameters include: the phase of the signal light, the green light window time, and the signal cycle; the adaptive signal control algorithm is a reinforcement learning-based control algorithm that can dynamically generate preventive control strategies for signal lights at waterway junctions.

[0078] After acquiring comprehensive traffic situation information and traffic prediction information, an adaptive signal control algorithm based on reinforcement learning is adopted. A state space S is defined to construct a complete perception dimension encompassing "real-time state, predicted trends, and situational characteristics." The design goal of state space S is to enable the reinforcement learning agent to comprehensively and accurately perceive the current state, future trends, and differentiated characteristics of waterway traffic, providing sufficient basis for action decisions. An action space A is also defined, where actions are a set of adjustment instructions for phases and durations of different types of vessels. A reward function R is defined, aiming to minimize the predicted total waiting time of all vessels approaching the junction within a future signal cycle and maximize the predicted passage efficiency of the junction. The agent then adjusts its actions based on the current state... Select Action Execution, environment state transfer to And generate rewards Through continuous iteration and optimization of strategies In order to obtain the optimal signal control strategy.

[0079] In one embodiment, a specific implementation method is also provided for generating preventive control parameters for signal lights at key waterway nodes using an adaptive signal control algorithm based on comprehensive traffic situation information and traffic prediction information, including: A state space is constructed, using the traffic flow at key channel nodes at the current moment, average vessel waiting time, predicted congestion index, current signal light status, and the proportion of target vessels as state variables. An action space is also constructed, including phase adjustment actions and duration adjustment actions. Phase adjustment actions refer to instructions to switch from the currently operating signal phase to the next phase in a preset phase sequence, while duration adjustment actions refer to instructions to increase, decrease, or maintain the green light duration by a fixed time step based on the current phase. Based on predicted traffic flow, the predicted number of vessels passing through key channel nodes in the next signal cycle is estimated. Based on predicted arrival rate, predicted average speed, and the remaining waiting time of vessels currently in the waiting queue, the predicted total waiting time for all vessels in the next signal cycle is calculated. A reward function is constructed by weighting the predicted number of vessels passing through and the predicted total waiting time. Based on the state space, action space, and reward function, and following the criteria of minimizing the predicted total waiting time and maximizing predicted traffic efficiency, a deep Q-network algorithm is used to iteratively optimize the initial control parameters in the deep Q-network to obtain preventative control parameters.

[0080] The average waiting time for vessels mentioned above is calculated based on comprehensive traffic situation information, taking into account the waiting time of all vessels currently waiting in the queue at the traffic lights. The current traffic light status includes the current phase of the traffic light and the remaining time of the current phase. The proportion of target vessels is the proportion of vessels whose tonnage exceeds a preset threshold and which meet the target vessel type, based on the vessel type and tonnage information in the dynamic and static attribute information of all vessels.

[0081] Specifically, the aforementioned state space includes state variables. The selection of these state variables balances comprehensiveness and specificity. These state variables include: real-time traffic flow, average vessel waiting time, predicted congestion index, and current traffic light status. The real-time traffic flow... This refers to the real-time traffic flow of vessels at key nodes in the waterway (such as junctions and lock entrances), reflecting the immediate level of congestion and serving as the basis for determining whether signal adjustments are necessary. Average vessel waiting time. Based on the acquired digital panoramic view of waterway traffic, the system calculates the arithmetic mean of the waiting times of all vessels currently queuing at traffic lights, quantifying the immediate effect of the current signaling strategy. Longer waiting times indicate a lower degree of alignment between the current strategy and traffic demand. This predicts the congestion index. This is the congestion index for the next predicted time slice, allowing the intelligent agent to perceive future traffic trends in advance, providing core support for preventative control and avoiding reactive scheduling. Current traffic light status. Includes the current signal phase and the remaining time of the current phase This ensures that action decisions are seamlessly integrated with the existing signal status, avoiding navigation chaos caused by phase switching conflicts or sudden changes in duration.

[0082] The state space S described above can also incorporate ship composition features, such as the proportion of the target ship. To adapt to differentiated traffic flow compositions, the proportion of target vessels is based on the vessel type in the dynamic and static attribute information of the vessels. and tonnage information The target vessel can be a large vessel that meets the standard conditions for large vessels. Based on the vessel type and tonnage information, the proportion of large vessels with tonnage exceeding a preset threshold among the vessels currently approaching the junction is calculated. This variable is used to make the preventive control parameters more adaptable to traffic flows composed of vessels with different standard conditions.

[0083] The proportion of target ships state space In this approach, the characteristics of vessel composition are used as state variables, enabling the reinforcement learning agent to perceive differences in traffic flow composition. For example, when the proportion of large vessels is high, the algorithm may automatically extend the green light time to accommodate their slow start-up and long passage time, thereby significantly improving the precision and practicality of the control strategy. By deeply integrating predictive information and incorporating traffic flow composition characteristics, the adaptive signal control algorithm acquires forward-looking and context-aware capabilities.

[0084] After constructing the state space, define the action space. The design of this motion space focuses on the core operational dimensions of signal control. Through a combination of "phase adjustment" and "duration adjustment," it achieves precise adaptation to different traffic scenarios, and its definition balances feasibility and flexibility. Motion space A includes phase adjustment actions. and duration adjustment action Phase adjustment refers to the instruction to switch from the currently operating signal phase to the next phase in a preset phase sequence. For example, switching from "upstream vessels have priority passage phase" to "downstream vessels have priority passage phase." Its core function is to allocate passage rights for vessels of different directions and priorities, avoiding one-way congestion caused by a single phase lasting too long. Duration adjustment refers to the instruction to "increase, decrease, or keep unchanged" the green light duration by a fixed time step Δt, while maintaining the current phase. Here, Δt is a fixed time step, such as 10 seconds, which can be configured according to the channel's navigation efficiency requirements. The core function of this action is to dynamically adapt to traffic flow changes, extending the green light duration during peak traffic periods to increase throughput, and shortening the green light duration during off-peak traffic periods to reduce waiting time in other directions. (Complete action...) It consists of a combination of phase adjustment actions and duration adjustment actions. For example: "Switch to the phase for outgoing vessels + increase the green light duration by 10 seconds" or "Keep the current phase for incoming vessels + the green light duration unchanged". This combination mode covers the control needs of all scenarios, including smooth traffic, light congestion, and heavy congestion, ensuring the flexibility and executability of the strategy.

[0085] Furthermore, based on the above reinforcement learning-based control algorithm, a reward function is constructed. The reward function R is the "navigation criterion" of the reinforcement learning agent. Its core design is to transform the business objective of reducing ship waiting time and improving waterway traffic efficiency into a quantifiable mathematical indicator. The formula for the reward function R can be expressed as follows: ; in, This is the predicted total waiting time for all vessels approaching key nodes (junctions) in the channel within the next signal cycle. To predict the number of passing vessels, and These are the weighting coefficients, and .

[0086] Taking a key waterway node as an example of a junction, determining the predicted total waiting time can be based on predicted traffic conditions. This involves determining the remaining waiting time of existing vessels already in the queue, which can be understood as the predicted total waiting time for vessels arriving at the junction within a future period. It also includes the expected waiting time for new vessels joining the queue within the future period, which can be understood as vessels estimated to join the queue within the future period based on predicted arrival rates and predicted average speeds. The aforementioned weighting coefficients can be dynamically adjusted according to waterway management priorities. For example, when prioritizing navigation experience, the weighting coefficients can be increased. Increase when focusing on freight efficiency Among them, the predicted number of ships passing through the junction in the next cycle is directly related to the channel's traffic efficiency. The higher this value, the higher the reward value, which guides the agent to improve the overall traffic capacity of the channel.

[0087] After constructing the state space, action space, and reward function, and following the principles of minimizing the total predicted waiting time and maximizing the predicted passage efficiency, the initial control parameters in the deep Q-network are iteratively optimized using the deep Q-network algorithm to obtain the preventative control parameters. Specifically, the deep Q-network is first initialized. The network, based on the current state in the state space, selects and executes the current action from the action space according to the initial control parameters, obtains an immediate reward, and transitions the environment state from the current state to the next state. Through an experience replay mechanism, transition samples generated from each interaction are stored in the experience replay pool. Each transition sample includes the current state, the current action, the immediate reward, and the next state. When training conditions are met, random batches of transition samples are sampled from the experience replay pool, and the corresponding target Q-value is calculated. Following the criteria of minimizing the total predicted waiting time and maximizing the predicted passage efficiency, the parameters of the deep Q-network are updated using gradient descent based on the target Q-value, and the initial control parameters are optimized until the deep Q-network converges, yielding preventative control parameters.

[0088] In this embodiment, a Deep Q-Network (DQN) algorithm is used to iteratively optimize the reinforcement learning policy, solving the problems of action value estimation and policy convergence in high-dimensional state spaces, and ensuring the stability and reliability of the generated control policy. The Deep Q-Network is a Q-function approximator, and its input is the current state. The output is for each optional action. goal Value; stored transfer samples through an experience replay mechanism ( , , , The network uses random sampling to train on the data, breaking down correlations between data points; it also uses a target network to provide a stable target. The value is periodically synchronized with the online network to improve training stability. The target Q-value represents "in the current state..." Next action The agent achieves optimal decision-making by selecting the action that maximizes the target Q-value, thus obtaining long-term cumulative rewards.

[0089] Specifically, please see Figure 8 As shown, starting from the initiation of adaptive signal control for the waterway and the initialization of the reinforcement learning agent (e.g., DQN network), the interaction loop between the agent and the waterway environment begins. The agent first monitors the current state of the waterway. Based on the current state, the corresponding traffic light control action is output according to the initial control parameters. This means issuing a signal light adjustment command; after the action is executed, the environment simulator / navigation environment will provide feedback on the updated next state. At the same time, the reward corresponding to this action is calculated based on the next state. For example, congestion relief and improved traffic efficiency correspond to positive rewards, while increased congestion corresponds to negative rewards; the transition samples generated from each interaction ( , , , The data is stored in the experience replay pool. When the preset training rounds are reached, the system extracts random batches of training data from the experience replay pool, including information such as state, action, and reward. This breaks the correlation between data, avoids model training oscillations, and improves training stability. The system then sequentially calculates the target Q-value, updates the online Q-network weights using gradient descent, and periodically updates the target network to optimize the initial control parameters. After optimizing the agent's parameters, the system enters the next interaction loop (t=t+1), repeating the above process iteratively until the model converges, yielding preventative control parameters. The target network is used to provide a stable target... The values ​​are periodically synchronized with the online network to improve training stability.

[0090] The generation period of the preventive control parameters in the adaptive signal control strategy is related to the future preset time window in the above embodiments. To maintain synchronization, each control decision is based on the latest and most forward-looking traffic condition predictions. For example, if the traffic condition is predicted for the next 30 minutes (K=6, time slice interval of 5 minutes) in the above embodiment, a new control strategy is generated every 5 minutes in this embodiment to ensure that each decision is based on the latest prediction results, achieving a closed-loop linkage of "prediction-decision-execution".

[0091] This embodiment introduces a predicted congestion index and a reward function to fuse predicted waiting time and traffic volume in the state space, achieving preventative control and avoiding congestion in advance. Furthermore, it incorporates the proportion of large vessels into the state space, allowing the strategy to perceive differences in traffic flow composition and adapt to the passage needs of different types of vessels, improving situational adaptability. The strategy also achieves autonomous iterative optimization based on the DQN algorithm, eliminating the need for manually preset control rules and adapting to long-term changes in waterway traffic flow. By introducing reinforcement learning algorithms and continuously interacting with the waterway environment to achieve self-iterative optimization, it can generate optimal control strategies for complex traffic scenarios, overcoming the decision-making limitations of fixed-rule or simple-response systems. Simultaneously, by incorporating vessel composition (Rlarge) and meteorological data (Wspeed, Visibility) into the decision-making basis, the control strategy fully adapts to actual operating conditions, significantly improving the robustness and accuracy of decision-making. Ultimately, compared to traditional systems, this scheme gives the control strategy stronger adaptability and a higher level of intelligence, dynamically responding to complex and changing waterway traffic environments, ensuring that signal control and vessel guidance always meet actual needs, and effectively improving the stability and efficiency of waterway operation.

[0092] Step S205: Based on traffic forecast information and preventive control parameters, generate traffic guidance information for each vessel, send the traffic guidance information to the corresponding vessel terminal, and send the preventive control parameters to the corresponding signal light controller.

[0093] The process of generating traffic guidance information can be broken down into several sub-steps, including "two modes of speed guidance (on-time arrival / smooth deceleration) → speed calculation under safety constraints → prediction-based path cost assessment → multi-objective path selection → collaborative allocation strategy", which provides ships with individualized and globally optimal passage solutions, achieving efficient passage for individual ships while ensuring load balance across the entire waterway network.

[0094] Traffic guidance information includes speed guidance information and route recommendation information; speed guidance information includes a first recommended speed or a second recommended speed; route recommendation information includes alternative recommended routes. The first recommended speed is the recommended speed corresponding to the on-time arrival mode, and the second recommended speed is the recommended speed corresponding to the smooth deceleration mode.

[0095] Based on traffic forecast information and preventative control parameters, traffic guidance information is generated for each vessel, including: Based on the predicted congestion index, the speed guidance mode for ships at key nodes is determined. The speed guidance mode includes the on-time arrival mode and the smooth deceleration mode. The on-time arrival mode refers to the mode in which the predicted congestion index does not exceed the preset index threshold and the green light passage is guaranteed at the key node. The smooth deceleration mode refers to the mode in which the predicted congestion index exceeds the preset index threshold and the ship cannot be guaranteed to pass through the next green light window.

[0096] When in on-time arrival mode, the real-time distance between the vessel and key nodes is acquired. Based on the real-time distance and the green light window time in the preventive control parameters, a first recommended speed is calculated for the vessel to ensure that the vessel's expected arrival time falls within the green light window time. The first recommended speed is greater than the minimum sailing speed but does not exceed the upper limit of the safe speed. The upper limit of the safe speed is calculated based on the vessel's static attribute information and channel conditions using a vessel tonnage-speed matching model. When in smooth deceleration mode, a second recommended speed is calculated for the vessel. The second recommended speed is used to allow the vessel to approach the congested area in a smooth deceleration manner. The second recommended speed is less than the first recommended speed. The second recommended speed is greater than the minimum sailing speed but does not exceed the upper limit of the safe speed.

[0097] Based on the predicted average speed and waterway network topology data for each waterway segment, a graph theory algorithm is used to construct a cost function for each node in the waterway network topology data, and multiple candidate paths are output according to the cost function. The cost function includes the actual cost from the origin to the node and the estimated cost from the node to the destination. The actual cost is calculated based on the predicted average speed and the length of the path segment, while the estimated cost is estimated based on the predicted travel time. Based on the predicted total travel time of the candidate path, the average predicted congestion index of each waterway segment on the candidate path, and the path complexity, a comprehensive score is calculated for each candidate path through a multi-objective evaluation function. The candidate path with the highest comprehensive score is selected as the optimal recommended path in descending order of comprehensive score. The path complexity is calculated by weighting the number of key nodes and the number of turns on the candidate path. Based on the optimal recommended path and the traffic load data of the entire waterway, the congestion risk of multiple ships choosing the same path is assessed, and the assessment results are obtained. Based on the assessment results, ships are dynamically assigned to different alternative paths, generating alternative recommended paths.

[0098] Specifically, based on the predicted congestion index, the speed guidance mode for ships at key nodes is determined. When the predicted congestion index does not exceed the preset index threshold, it indicates that the ship is in the on-time arrival mode; when the predicted congestion index exceeds the preset index threshold, it indicates that the ship is in the smooth deceleration mode.

[0099] Taking a key node as an example, when in on-time arrival mode, for ships Obtain the real-time distance between it and the downstream target intersection. Obtain the predicted congestion index at the intersection. And preventative control parameters, which include the green light window time for the traffic lights for the next few cycles. ; and obtain the planned sailing time and calculate the ship's speed based on the planned sailing time and real-time distance. First recommended speed The calculation formula is as follows: ; in, The planned sailing time aims to make the ship Estimated arrival time Falling within the green light window period, that is ≤ ≤ , For ships Real-time distance to the intersection with the downstream target .

[0100] When in smooth deceleration mode, i.e. when continuous congestion is predicted at the junction and green light passage cannot be guaranteed, the adjustment target is to calculate a lower second recommended speed to guide ships to approach the congestion area in a smooth deceleration manner, avoiding abrupt stops and queuing, while reducing fuel consumption. This degradation strategy ensures that the system can still maintain navigation order in extreme congestion scenarios, avoid abrupt stops and dense queues at the junction, and minimize fuel consumption. The adjustment method clearly defines the degradation optimization strategy when "green wave passage" is not possible, demonstrating the robustness of the system.

[0101] It should be noted that when calculating the first and second recommended speeds, the ship's position must also be taken into account. The static properties are constrained to ensure that the recommended speed simultaneously meets both types of constraints, thus preventing overspeeding. In other words, the recommended speed must not exceed the safe speed limit. And not lower than the minimum sailing speed stipulated for that section of the waterway. Among these methods, based on the ship's tonnage information and waterway conditions, the upper limit of the ship's safe speed is calculated using a "tonnage-speed matching model". For example, large vessels, due to their poor maneuverability, have a lower upper limit for safe speed; the speed must not be lower than the minimum speed stipulated for a section of the waterway to avoid affecting the overall traffic efficiency of the waterway. These two types of constraints ensure that the recommended speed is both suitable for the characteristics of the vessel and complies with the waterway safety rules.

[0102] Furthermore, the predicted average speed and channel network topology data for each channel segment are obtained, and an improved A... Algorithm (A-Star) calculation The core of this candidate path approach is to define a cost function by predicting travel time, ensuring that the path adapts to future traffic conditions. A graph theory algorithm is used to construct the cost function for each node in the waterway network topology data. .in, From the starting point to the node The actual cost, i.e. the predicted travel time from the starting point to the current node, can be calculated from the predicted average speed of the path segment and the path length. For the node The estimated cost to the destination, i.e., the predicted travel time from the current node to the destination, is also based on the predicted speed estimate. This is calculated by... (The sentence is incomplete and requires more context to translate accurately.) Total cost The algorithm is guided to prioritize exploring nodes that are "most likely to be close to the optimal path," that is, to prioritize exploring paths with shorter future travel times as candidate paths, and to avoid recommending channel segments that have been predicted to be congested.

[0103] Based on the predicted total travel time of the candidate paths, the average predicted congestion index of each waterway segment on the candidate paths, and the path complexity, a multi-objective evaluation function is used to evaluate each candidate path. Calculate the overall score It can be expressed by the following formula: ; in, Candidate paths The predicted total travel time; Candidate paths The average predicted congestion index for each waterway segment; Candidate paths The complexity is calculated by weighting the number of intersections and turns on the path; , , These are the weighting coefficients for each indicator, which can be adjusted according to waterway management needs.

[0104] Based on the comprehensive score The candidate paths are ranked, and the candidate path with the highest score is selected as the optimal recommended path. In this embodiment, a quantitative multi-objective evaluation function is used so that the path recommendation no longer relies solely on the shortest distance or time, but comprehensively considers congestion and navigation difficulty, thus making the recommendation results more scientific and user-friendly.

[0105] Optionally, before sending alternative recommended routes to vessels, a collaborative allocation operation is included. This operation assesses the congestion risk of multiple vessels choosing the same route based on the optimal recommended route and overall waterway traffic load data. If more than a preset number of vessels simultaneously receive the same route recommendation, new congestion will occur on that route. Based on the assessment results, the recommendation strategy for the optimal route is fine-tuned, and the vessel group is collaboratively allocated to different alternative routes to achieve dynamic load balancing across the entire waterway network. Through collaborative allocation, the "congestion shifting" problem caused by blind recommendations can be avoided, achieving true dynamic network load balancing optimization.

[0106] When the aforementioned speed guidance information and route recommendation information are sent to the corresponding ship terminals through the transit network, they can be displayed graphically on the electronic river chart on the ship terminals, and crew members can be provided with an interactive interface for one-click confirmation or modification.

[0107] In this embodiment, on the one hand, by generating speed guidance information and path recommendation information, individualized speed guidance suggestions can be issued to each vessel in the waterway. This not only helps vessels achieve "green wave passage" to improve traffic efficiency, but also guides vessels to smoothly decelerate before congested areas, reducing sudden stops and starts, effectively reducing unnecessary fuel consumption and exhaust emissions, thus balancing economy and environmental protection. On the other hand, through a collaborative path recommendation strategy, the traffic load of the entire waterway network is balanced at the spatial level, completely avoiding congestion transfer problems and ensuring smooth passage at each node of the waterway. Ultimately, the combined effect of the above dual optimization methods achieves overall smooth traffic flow, significantly improves the overall operating efficiency of the waterway network, and solves the problems of low traffic efficiency and resource waste caused by disorderly vessel passage and congestion transfer in the traditional mode.

[0108] In one embodiment, traffic guidance information is sent to the corresponding vessel terminal, and preventative control parameters are sent to the corresponding signal light controller, including: Preventive control parameters are encapsulated into standard control messages, and corresponding standard information messages are generated for each vessel based on traffic guidance information. The standard control messages and standard information messages are sorted according to information transmission priority, and the sorting result is obtained. The priority of the standard control messages is higher than that of the standard information messages. Based on the sorting result, the standard control messages are sent to the signal light controllers at key nodes in the waterway, and the target messages and other messages to be compressed are determined from the standard information messages. The other messages are sent to the corresponding vessel terminals in standard format, and the target messages are transmitted to the corresponding vessel terminals after data compression processing.

[0109] Specifically, the aforementioned preventative control parameters include phase commands. Green light duration and signal period The system encapsulates the signal into a standard control message according to a preset communication protocol and sends the standard control message to the signal controller at the target channel junction via a wireless communication network, in accordance with the communication protocol. This allows the signal controller to receive and parse the standard control message, and then perform corresponding control operations on the signal lights according to the preventive control parameters, thereby driving the signal lights to switch to the specified state.

[0110] Based on traffic guidance information, a corresponding standard information message is generated for each vessel. This standard information message may include a recommended speed and an alternative recommended route, as well as a valid timestamp for the standard information message. Standard information messages are transmitted to ships via wireless communication networks in unicast or multicast mode. The corresponding ship terminal.

[0111] The wireless communication network can be a 5G communication network or a VHF data exchange system. During the transmission of standard control messages and standard information messages, transmission can be performed on uRLLC (Ultra-Reliable Low-Latency Communication) slices of the 5G network or specific data channels of VHF, based on data priority requirements, to ensure transmission reliability and real-time performance. Standard control messages have a higher priority than standard information messages. Standard control messages can be sent first to the signal light controllers at key waterway nodes, and target messages and other messages to be compressed can be determined from the standard information messages. Other messages are those with higher urgency, while target messages are those with lower urgency. Urgency can be determined based on the distance between the vessel and the congestion point.

[0112] On the ship's terminal, traffic guidance information can be presented in a differentiated manner through a human-computer interaction interface. For example, the recommended speed can be displayed in a striking numerical form, supplemented by arrows indicating acceleration and deceleration; route recommendation information can be superimposed on the visualization interface of the electronic river chart with highlighted or different colored lines; and for some collaborative allocation suggestions, brief explanatory information can be provided to the crew.

[0113] Before sending information, all instructions and information to be sent (including standard control messages and standard information messages) are prioritized and compressed. Signal light control instructions are given the highest priority and sent immediately to the signal light controller. Speed ​​guidance information and route recommendation information are prioritized and sent according to their urgency. Non-urgent target messages can be compressed using data compression algorithms to reduce network load. Other messages are sent to the corresponding ship terminals in standard format. This step, by prioritizing and compressing data, addresses potential communication bandwidth limitations in inland waterways. Through intelligent scheduling and compression technology, the timeliness of critical control instructions is ensured, improving the overall robustness and usability of the system.

[0114] After the ship terminal receives the standard information message, a feedback and confirmation mechanism is also included: the ship terminal returns a reception confirmation signal to the data processing and control center. If a crew member operates through the terminal interface and generates a corresponding confirmation execution command or modification command, the ship's terminal will also send the crew member's confirmation execution command. Or modify instructions Feedback is sent back to the data processing and control center. If the data processing and control center does not receive a reception confirmation signal from the ship's terminal module within a preset time... or received modification instructions When this happens, a message retransmission or policy update process will be triggered. It introduces the concepts of two-way communication and closed-loop control, ensuring the system knows whether the instruction has been delivered and accepted, greatly improving the system's reliability and intelligence.

[0115] This application provides a method for alleviating congestion in inland waterways. Compared with existing technologies, this solution collects multi-source traffic data from vessels and meteorological and hydrological data from the waterway, and generates comprehensive traffic situation information through spatiotemporal registration and fusion processing. This achieves a comprehensive perception of vessel distribution, movement status, identity attributes, and overall traffic conditions in the waterway, breaking through the limitations of related technologies that rely solely on radar monitoring of queue lengths. This provides comprehensive and accurate baseline data for waterway operation for subsequent decision-making. Furthermore, by inputting traffic flow characteristic parameters, waterway congestion index, and meteorological and hydrological data into a trained prediction model, it can obtain the traffic status at key nodes within a preset time window in advance, possessing the ability to predict waterway traffic conditions proactively. This upgrades signal control from a reactive, post-event response to a proactive, preventative approach. To mitigate congestion, the system utilizes adaptive signal control algorithms to generate preventative control parameters for traffic lights based on comprehensive traffic conditions and forward-looking traffic forecasts. This move moves beyond reactive adjustments to green light durations after congestion, proactively optimizing signal parameters to anticipate potential congestion and prevent further congestion. Furthermore, by combining traffic forecasts and preventative control parameters, the system generates individualized traffic guidance information for each vessel and pushes it to the terminal, achieving two-way coordination between traffic lights and proactive vessel guidance. This not only enables precise and intelligent control of traffic lights but also provides advance path and speed guidance for vessels. It addresses the issues of incomplete technical decision-making and the inability to provide advance guidance, which lead to increased congestion and low traffic efficiency. This significantly improves the overall traffic efficiency and operational stability of the waterway.

[0116] Based on the same inventive concept, this application also provides an inland waterway congestion relief device for implementing the aforementioned method. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more inland waterway congestion relief device embodiments provided below can be found in the limitations of the inland waterway congestion relief method described above, and will not be repeated here.

[0117] In one exemplary embodiment, such as Figure 9 As shown, an inland waterway congestion relief device is provided, the device comprising: The acquisition module 510 is used to acquire multi-source traffic data and meteorological and hydrological data of various vessels in the waterway. The processing module 520 is used to perform spatiotemporal registration and fusion processing on multi-source traffic data to generate comprehensive traffic situation information for the entire waterway. Based on the comprehensive traffic situation information, traffic flow characteristic parameters and waterway congestion index are extracted. The comprehensive traffic situation information is used to characterize the distribution location, movement status, identity attributes, and overall waterway traffic status of all ships in the waterway under a unified coordinate system. The waterway congestion index is used to reflect the overall smoothness of waterway traffic. The prediction module 530 is used to input traffic flow characteristic parameters, waterway congestion index and waterway meteorological and hydrological data into the trained traffic flow prediction model to obtain traffic prediction information of key nodes in the waterway in the future preset time window. The generation module 540 is used to generate preventive control parameters for signal lights at key nodes of the waterway based on comprehensive traffic situation information and traffic prediction information using an adaptive signal control algorithm. The sending module 550 is used to generate traffic guidance information for each vessel based on traffic forecast information and preventive control parameters, send the traffic guidance information to the corresponding vessel terminal, and send the preventive control parameters to the corresponding signal light controller.

[0118] As an optional implementation, the processing module 520 is specifically used for: Establish a waterway geographic coordinate system and convert the first position information, the second position information, and the field of view information of the camera into the waterway geographic coordinate system; A unified timestamp is established for AIS information, ship visual information, and target detection information. Based on the timestamp, asynchronous data in multi-source traffic data is time-synchronized and aligned for time registration. After time registration is completed, the first position information and velocity vector are used as observations, and the second position information and ground speed are used as auxiliary correction quantities. A data fusion algorithm is used for fusion filtering to obtain the fused trajectory data of each ship. The fused trajectory data includes: the optimal estimated position and the optimal estimated speed. The fusion trajectory data is linked and bound with identity identifiers and static attribute information to assign dynamic and static attribute information to each ship; After the attributes are assigned, a digital panoramic view of waterway traffic is constructed in a unified coordinate system based on the fused trajectory data and dynamic and static attribute information of all ships.

[0119] As an optional implementation, the processing module 520 is also used for: Within a preset time interval, the number of vessels at key sections of the waterway is counted, and the traffic flow per unit time is calculated based on the number of vessels. The average speed of the critical area is obtained by arithmetically averaging the optimal estimated speeds of all ships within the critical area of ​​the waterway. At a preset time, the sum of the projected lengths of all ships in the channel segment and the total length of the channel segment are obtained, and the ratio of the sum of the projected lengths to the total length of the channel segment is processed to obtain the channel occupancy rate. Obtain the total number of ships and area of ​​key regions, and calculate the ship density based on the total number of ships and area; The waterway congestion index is obtained by weighting and integrating traffic flow, average speed, waterway occupancy, and vessel density.

[0120] As an optional implementation, the prediction module 530 is specifically used for: The traffic flow characteristic parameters and waterway congestion index of the past T time slices of key nodes in the waterway are obtained by normalization preprocessing to obtain the processed data; T and K are both positive integers. The processed data and corresponding waterway meteorological and hydrological data are used as input sequences. Through a trained traffic flow prediction model, the predicted parameters are output and inversely normalized to obtain the traffic flow, predicted average speed, and predicted congestion coefficient for the next K time slices. The traffic flow prediction model consists of an input layer, at least one hidden layer, and a fully connected output layer connected in sequence. The input layer is used to receive the input sequence, the hidden layer is used to learn the long-term and short-term dependencies in the traffic flow feature parameters, and the fully connected output layer is used to map the final output of the hidden layer into traffic prediction information. The traffic prediction information is represented by a traffic prediction sequence.

[0121] As an optional implementation method, the traffic flow prediction model is constructed through the following steps: Historical waterway traffic flow characteristic data is obtained, and the historical waterway traffic flow characteristic data is normalized and preprocessed to obtain processed historical data. The processed historical data is divided into training and validation sets according to a preset ratio; the processed historical data includes real sequences. The processed historical data is input into the initial model for time series prediction, and the predicted sequence is output. Based on the predicted sequence and the true sequence, a loss function is constructed. By minimizing the loss function, an optimizer is used to iteratively update the parameters in the initial model to obtain the trained model. The validation set is used to validate the performance of the trained model in order to build a traffic flow prediction model.

[0122] As an optional implementation, the generation module 540 is specifically used for: A state space is constructed, with the traffic flow at key nodes of the waterway at the current moment, the average waiting time of ships, the predicted congestion index, the current signal light status, and the proportion of target ships as state variables. The average waiting time of ships is calculated based on comprehensive traffic situation information, taking into account the waiting time of all ships currently waiting in the queue in front of the signal light. The current signal light status includes the current phase of the signal light and the remaining time of the current phase. The proportion of target ships is determined based on the ship type and tonnage information in the dynamic and static attribute information of all ships, and is the proportion of the number of ships whose tonnage exceeds a preset threshold and matches the target ship type out of the total number of ships. Construct an action space; the action space includes phase adjustment actions and duration adjustment actions; a phase adjustment action is an instruction to switch from the currently running signal phase to the next phase in the preset phase sequence, and a duration adjustment action is an instruction to increase, decrease, or maintain the green light duration by a fixed time step based on the current phase; Based on the predicted traffic flow, estimate the predicted number of vessels passing through key nodes of the waterway in the next signal cycle. Based on the predicted arrival rate, predicted average speed, and remaining waiting time of vessels currently in the waiting queue, the predicted total waiting time of all vessels in the next signal cycle is calculated. The predicted arrival rate is determined based on traffic forecast information. All vessels include existing vessels currently in the waiting queue and new vessels to be added in the future cycle. New vessels are those that will join the waiting queue in the future cycle based on the predicted arrival rate and predicted average speed. A reward function is constructed by weighting the predicted number of passing vessels and the predicted total waiting time. Based on the state space, action space, and reward function, and following the principles of minimizing the total predicted waiting time and maximizing the predicted passage efficiency, the initial control parameters in the deep Q-network are iteratively optimized using the deep Q-network algorithm to obtain the preventive control parameters.

[0123] As an optional implementation, the generation module 540 is also used for: Initial depth The network, based on the current state in the state space, selects and executes the current action from the action space according to the initial control parameters, obtains an immediate reward, and the environment state transitions from the current state to the next state; The experience replay mechanism stores the transition samples generated by each interaction into the experience replay pool; the transition samples include the current state, the current action, the immediate reward, and the next state. When the training conditions are met, random batches of transfer samples are sampled from the experience replay pool, and the corresponding target Q value is calculated. Based on the principles of minimizing the total predicted waiting time and maximizing the predicted passage efficiency, the parameters of the deep Q-network are updated using gradient descent according to the target Q-value, and the initial control parameters are optimized until the deep Q-network converges, thus obtaining the preventive control parameters.

[0124] As an optional implementation, the sending module 550 is specifically used for: Based on the predicted congestion index, the speed guidance mode for ships at key nodes is determined. The speed guidance mode includes the on-time arrival mode and the smooth deceleration mode. The on-time arrival mode is the mode in which the predicted congestion index does not exceed the preset index threshold and the green light passage is guaranteed at the key node. The smooth deceleration mode is the mode in which the predicted congestion index exceeds the preset index threshold and the ship cannot be guaranteed to pass through the next green light window. When in on-time arrival mode, the real-time distance between the vessel and key nodes is obtained, and a first recommended speed is calculated for the vessel based on the real-time distance, the green light window time in the preventive control parameters, so that the vessel's expected arrival time falls within the green light window time; the first recommended speed is greater than the minimum sailing speed but does not exceed the upper limit of safe speed; the upper limit of safe speed is calculated based on the vessel's static attribute information and waterway conditions through the vessel tonnage-speed matching model; When in smooth deceleration mode, a second recommended speed is calculated for the vessel. The second recommended speed is used to allow the vessel to approach the congested area in a smooth deceleration manner. The second recommended speed is less than the first recommended speed. The second recommended speed is greater than the minimum sailing speed but does not exceed the upper limit of the safe speed. Based on the predicted average speed and channel network topology data of each channel segment, a graph theory algorithm is used to construct the cost function of each node in the channel network topology data, and output multiple candidate paths according to the cost function. The cost function includes the actual cost from the starting point to the node and the estimated cost from the node to the destination. The actual cost is calculated based on the predicted average speed and the path segment length, and the estimated cost is estimated based on the predicted travel time. Based on the predicted total travel time of the candidate path, the average predicted congestion index of each channel segment on the candidate path, and the path complexity, a comprehensive score is calculated for each candidate path using a multi-objective evaluation function. The candidate path with the highest comprehensive score is selected as the optimal recommended path in descending order of comprehensive score. The path complexity is calculated by weighting the number of key nodes and the number of turns on the candidate path. Based on the optimal recommended route and the traffic load data of the entire waterway, the congestion risk of multiple ships choosing the same route is assessed, and the assessment results are obtained. Based on the assessment results, ships are dynamically assigned to different alternative routes, and alternative recommended routes are generated.

[0125] As an optional implementation, the sending module 550 is also used for: Preventive control parameters are encapsulated into standard control messages, and corresponding standard information messages are generated for each vessel based on traffic guidance information; The standard control messages and standard information messages are sorted according to their information transmission priority, and the sorting result is obtained; the priority of the standard control message is higher than that of the standard information message. Based on the sorting results, standard control messages are sent to the signal light controllers at key nodes in the waterway, and target messages and other messages to be compressed are determined from the standard information messages. Other messages are sent to the corresponding ship terminals in standard format, and the target message is transmitted to the corresponding ship terminals after data compression processing.

[0126] The inland waterway congestion mitigation device provided in this application collects multi-source traffic data from vessels and meteorological and hydrological data from the waterway. Through spatiotemporal registration and fusion processing, it generates comprehensive traffic situation information, achieving a complete perception of vessel distribution, movement status, identity attributes, and overall traffic conditions. This overcomes the limitations of related technologies that rely solely on radar monitoring of queue lengths, providing comprehensive and accurate baseline data for subsequent decision-making. Furthermore, by inputting traffic flow characteristic parameters, waterway congestion index, and meteorological and hydrological data into a trained prediction model, it can obtain the traffic status at key nodes within a preset time window in advance, possessing a forward-looking predictive capability for waterway traffic conditions. This upgrades signal control from a reactive, post-event response to a proactive, preventative control mechanism. To avoid exacerbating congestion, based on comprehensive traffic conditions and forward-looking traffic forecast information, an adaptive signal control algorithm is used to generate preventive control parameters for traffic lights. This is no longer limited to the delayed operation of "adjusting green light duration after congestion," but rather optimizes signal parameters in advance by predicting potential congestion, thus preventing congestion from escalating at the source. Furthermore, by combining traffic forecast information and preventive control parameters, individualized traffic guidance information is generated for each vessel and pushed to the terminal, achieving two-way coordination between traffic lights and proactive vessel guidance. This not only enables precise and intelligent control of traffic lights but also provides advance path and speed guidance for vessels. It solves the problems of congestion exacerbation and low traffic efficiency caused by one-sided technical decision-making and the inability to provide advance guidance, significantly improving the overall traffic efficiency and operational stability of the waterway.

[0127] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 10As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores video tag processing data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a method for alleviating congestion in inland waterways.

[0128] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0129] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0130] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0131] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0133] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0134] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0135] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0136] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for alleviating congestion in inland waterways, characterized in that, The methods for alleviating congestion in inland waterways include: Acquire multi-source traffic data and meteorological and hydrological data of various vessels in the waterway; The multi-source traffic data is spatiotemporally registered and fused to generate comprehensive traffic situation information for the entire waterway. Based on this comprehensive traffic situation information, traffic flow characteristic parameters and a waterway congestion index are extracted. The comprehensive traffic situation information is used to characterize the distribution location, motion state, identity attributes, and overall waterway traffic status of all vessels in the waterway under a unified coordinate system. The waterway congestion index is used to reflect the overall smoothness of waterway traffic. The traffic flow characteristic parameters, waterway congestion index, and waterway meteorological and hydrological data are input into the trained traffic flow prediction model to obtain traffic prediction information for key nodes in the waterway within a future preset time window. Based on the comprehensive traffic situation information and the traffic prediction information, an adaptive signal control algorithm is used to generate preventive control parameters for signal lights at key nodes in the waterway. Based on the traffic forecast information and the preventive control parameters, traffic guidance information is generated for each vessel, and the traffic guidance information is sent to the corresponding vessel terminal. The preventive control parameters are also sent to the corresponding signal light controller.

2. The inland waterway congestion relief method according to claim 1, characterized in that, The multi-source traffic data includes: AIS information, ship visual information, and target detection information; the target detection information includes the ship's first position information and velocity vector, and the AIS information includes the ship's second position information, ground speed, identification, and static attribute information; the comprehensive traffic situation information includes: a digital waterway traffic panorama. The multi-source traffic data is spatiotemporally registered and fused to generate comprehensive traffic situation information for the waterway, including: Establish a waterway geographic coordinate system, and convert the first position information, the second position information, and the field of view range information of the camera into the waterway geographic coordinate system; A unified timestamp is established for the AIS information, ship visual information, and target detection information. Based on the timestamp, asynchronous data in the multi-source traffic data is time-synchronized and aligned to perform time registration. After time registration is completed, the first position information and velocity vector are used as observations, and the second position information and the ground speed are used as auxiliary correction quantities. A data fusion algorithm is used for fusion filtering to obtain the fused trajectory data of each ship. The fused trajectory data includes: the optimal estimated position and the optimal estimated speed. The fused trajectory data is associated and bound with the identity identifier and static attribute information to assign dynamic and static attribute information to each ship; After the attributes are assigned, a digital panoramic view of waterway traffic is constructed in a unified coordinate system based on the fused trajectory data of all ships and the dynamic and static attribute information.

3. The inland waterway congestion relief method according to claim 2, characterized in that, The traffic flow characteristic parameters include: traffic volume, average speed, channel occupancy, and vessel density. Based on the comprehensive traffic situation information, key traffic flow characteristic parameters and waterway congestion index are extracted, including: Within a preset time interval, the number of vessels at key sections of the waterway is counted, and the traffic flow per unit time is calculated based on the number of vessels. The average speed of the critical area is obtained by arithmetically averaging the optimal estimated speeds of all ships within the critical area of ​​the waterway. At a preset time, the sum of the projected lengths of all ships in the channel segment and the total length of the channel segment are obtained, and the ratio of the sum of the projected lengths to the total length of the channel segment is processed to obtain the channel occupancy rate; Obtain the total number of ships and the area of ​​the key region, and calculate the ship density based on the total number of ships and the area. The traffic flow, average speed, channel occupancy, and vessel density are weighted and fused to obtain the channel congestion index.

4. The inland waterway congestion relief method according to claim 3, characterized in that, The traffic forecast information includes: predicted traffic flow, predicted average air speed, and predicted congestion index; the future specific time window includes K future time slices; The traffic flow characteristic parameters, waterway congestion index, and waterway meteorological and hydrological data are input into a trained traffic flow prediction model to obtain traffic prediction information for key nodes in the waterway within a future preset time window, including: The traffic flow characteristic parameters and waterway congestion index of the past T time slices of key nodes in the waterway are obtained by normalization preprocessing to obtain the processed data; T and K are both positive integers. The processed data and corresponding waterway meteorological and hydrological data are used as input sequences. A trained traffic flow prediction model is used to output prediction parameters, which are then inversely normalized to obtain the traffic flow, predicted average speed, and predicted congestion coefficient for the next K time slices. The traffic flow prediction model includes an input layer, at least one hidden layer, and a fully connected output layer connected in sequence. The input layer receives the input sequence, the hidden layer learns the long-term and short-term dependencies in the traffic flow feature parameters, and the fully connected output layer maps the final output of the hidden layer to the traffic prediction information, which is represented by a traffic prediction sequence.

5. The inland waterway congestion relief method according to claim 4, characterized in that, The traffic flow prediction model is constructed through the following steps: Historical waterway traffic flow characteristic data is obtained, and the historical waterway traffic flow characteristic data is preprocessed by normalization to obtain processed historical data; The processed historical data is divided into a training set and a validation set according to a preset ratio; the processed historical data includes real sequences; The processed historical data is input into the initial model for time series prediction, and the predicted sequence is output. Based on the predicted sequence and the true sequence, a loss function is constructed. By minimizing the loss function, an optimizer is used to iteratively update the parameters in the initial model to obtain the trained model. The validation set is used to validate the performance of the trained model in order to construct the traffic flow prediction model.

6. The inland waterway congestion relief method according to claim 4, characterized in that, The preventative control parameters include: the phase of the traffic light, the green light window time, and the signal cycle; Based on the comprehensive traffic situation information and the traffic prediction information, an adaptive signal control algorithm is used to generate preventative control parameters for signal lights at key waterway nodes, including: A state space is constructed, with the traffic flow at key nodes of the waterway at the current moment, the average waiting time of ships, the predicted congestion index, the current signal light status, and the proportion of target ships as state variables. The average waiting time of ships is calculated based on the comprehensive traffic situation information, taking into account the waiting time of all ships currently waiting in the queue at the signal light. The current signal light status includes the current phase of the signal light and the remaining time of the current phase. The proportion of target ships is determined based on the ship type and tonnage information in the dynamic and static attribute information of all ships, identifying the proportion of ships whose tonnage exceeds a preset threshold and that match the target ship type out of the total number of ships. Construct an action space; the action space includes phase adjustment actions and duration adjustment actions; the phase adjustment action refers to the instruction to switch from the currently running signal phase to the next phase in the preset phase sequence, and the duration adjustment action refers to the instruction to increase, decrease, or maintain the green light duration by a fixed time step based on the current phase; Based on the predicted traffic flow, estimate the predicted number of vessels passing through key nodes of the waterway in the next signal cycle. Based on the predicted arrival rate, predicted average speed, and remaining waiting time of the vessels currently in the waiting queue, the predicted total waiting time of all vessels in the next signal cycle is calculated; the predicted arrival rate is determined based on the traffic prediction information; all vessels refer to existing vessels currently in the waiting queue and newly added vessels in the future cycle, and the newly added vessels are those estimated to join the waiting queue in the future cycle based on the predicted arrival rate and predicted average speed. A reward function is constructed by weighting the predicted number of passing vessels and the predicted total waiting time. Based on the state space, the action space, and the reward function, and following the criteria of minimizing the total predicted waiting time and maximizing the predicted passage efficiency, the initial control parameters in the deep Q-network are iteratively optimized using the deep Q-network algorithm to obtain the preventive control parameters.

7. The inland waterway congestion relief method according to claim 6, characterized in that, Based on the state space, action space, and reward function, and following the criteria of minimizing the predicted total waiting time and maximizing the predicted passage efficiency, the initial control parameters in the deep Q-network are iteratively optimized using the deep Q-network algorithm to obtain the preventive control parameters, including: Initial depth The network, based on the current state in the state space, selects and executes the current action from the action space according to the initial control parameters to obtain an immediate reward, and the environment state transitions from the current state to the next state; The transition samples generated from each interaction are stored in the experience replay pool through an experience replay mechanism; the transition samples include the current state, the current action, the immediate reward, and the next state. When the training conditions are met, random batches of transfer samples are sampled from the experience replay pool, and the corresponding target Q value is calculated. Based on the criteria of minimizing the predicted total waiting time and maximizing the predicted passage efficiency, the parameters of the deep Q-network are updated using gradient descent according to the target Q-value, and the initial control parameters are optimized until the deep Q-network converges, thus obtaining the preventive control parameters.

8. The inland waterway congestion relief method according to claim 7, characterized in that, The traffic guidance information includes speed guidance information and route recommendation information; The speed guidance information includes a first recommended speed or a second recommended speed; The path recommendation information includes: alternative recommended paths; Based on the traffic forecast information and the preventive control parameters, traffic guidance information is generated for each vessel, including: Based on the predicted congestion index, the speed guidance mode for ships at the key nodes is determined; the speed guidance mode includes an on-time arrival mode and a smooth deceleration mode; the on-time arrival mode refers to the mode in which the predicted congestion index does not exceed a preset index threshold and green light passage is guaranteed at the key node; the smooth deceleration mode refers to the mode in which the predicted congestion index exceeds a preset index threshold and it cannot be guaranteed that the ship will pass through the next green light window. When in on-time arrival mode, the real-time distance between the vessel and key nodes is acquired, and a first recommended speed is calculated for the vessel based on the real-time distance, the green light window time in the preventive control parameters, so that the vessel's expected arrival time falls within the green light window time; the first recommended speed is greater than the minimum sailing speed and does not exceed the upper limit of safe speed; the upper limit of safe speed is calculated based on the vessel's static attribute information and channel conditions through a vessel tonnage-speed matching model; When in smooth deceleration mode, a second recommended speed is calculated for the vessel, which is used to allow the vessel to approach the congested area in a smooth deceleration manner; the second recommended speed is less than the first recommended speed; the second recommended speed is greater than the minimum sailing speed but does not exceed the upper limit of the safe speed. Based on the predicted average speed and channel network topology data of each channel segment, a graph theory algorithm is used to construct a cost function for each node in the channel network topology data, and multiple candidate paths are output according to the cost function. The cost function includes the actual cost from the starting point to the node and the estimated cost from the node to the destination. The actual cost is calculated based on the predicted average speed and the path segment length, and the estimated cost is estimated based on the predicted travel time. Based on the predicted total travel time of the candidate path, the average predicted congestion index of each channel segment on the candidate path, and the path complexity, a comprehensive score is calculated for each candidate path using a multi-objective evaluation function. The candidate path with the highest comprehensive score is selected as the optimal recommended path in descending order of comprehensive score. The path complexity is calculated by weighting the number of key nodes and the number of turns on the candidate path. Based on the optimal recommended route and the traffic load data of the entire waterway, the congestion risk of multiple ships choosing the same route is assessed to obtain the assessment results. Based on the assessment results, ships are dynamically assigned to different alternative routes to generate the alternative recommended routes.

9. The inland waterway congestion relief method according to claim 1, characterized in that, Sending the traffic guidance information to the corresponding ship terminal and sending the preventive control parameters to the corresponding signal light controller includes: The preventive control parameters are encapsulated into standard control messages, and corresponding standard information messages are generated for each vessel based on the traffic guidance information. The standard control messages and standard information messages are sorted according to their information transmission priority to obtain a sorting result; the priority of the standard control messages is higher than that of the standard information messages. Based on the sorting results, the standard control message is sent to the signal light controller of the key node of the waterway, and the target message and other messages to be compressed are determined from the standard information message. The other messages are sent to the corresponding ship terminals in a standard format, and the target message is transmitted to the corresponding ship terminals after data compression processing.

10. A congestion relief device for inland waterways, characterized in that, The inland waterway congestion relief device includes: The acquisition module is used to acquire multi-source traffic data and meteorological and hydrological data of various vessels in the waterway; The processing module is used to perform spatiotemporal registration and fusion processing on the multi-source traffic data to generate comprehensive traffic situation information for the entire waterway. Based on the comprehensive traffic situation information, traffic flow characteristic parameters and waterway congestion index are extracted. The comprehensive traffic situation information is used to characterize the distribution location, motion state, identity attributes, and overall waterway traffic status of all vessels in the waterway under a unified coordinate system. The waterway congestion index is used to reflect the overall smoothness of waterway traffic. The prediction module is used to input the traffic flow characteristic parameters, waterway congestion index and waterway meteorological and hydrological data into the trained traffic flow prediction model to obtain traffic prediction information of key nodes in the waterway in the future preset time window. The generation module is used to generate preventive control parameters for signal lights at key nodes of the waterway based on the comprehensive traffic situation information and the traffic prediction information using an adaptive signal control algorithm. The sending module is used to generate traffic guidance information for each vessel based on the traffic forecast information and the preventive control parameters, send the traffic guidance information to the corresponding vessel terminal, and send the preventive control parameters to the corresponding signal light controller.

Citation Information

Cited By

  • A port channel congestion monitoring and early warning method fusing multi-source heterogeneous data

    CN122176960A

  • A method for monitoring and early warning of port and waterway congestion by integrating multi-source heterogeneous data

    CN122176960B