Inland river transportation management system based on AI voyage planning

Abstract

The invention discloses an inland river transportation management system based on AI voyage planning, which relates to the technical field of intelligent scheduling of inland river transportation and comprises an information gathering module, an intelligent matching module, a manual auditing module and an AI voyage planning module. And the information aggregation module receives and stores the cargo source demand and the ship state information. And the intelligent matching module performs association analysis on the cargo source and the ship according to a preset rule in combination with the preliminary voyage evaluation to generate an initial matching scheme set. And the manual auditing module receives a dispatcher instruction, confirms the scheme and generates a transportation task. And the AI voyage planning module comprehensively calls inland waterway data and real-time environment data based on the confirmed transportation task, and generates a detailed voyage plan including a navigation path, a time node and energy consumption prediction through dynamic optimization calculation. According to the method, intelligent cooperation of matching and planning links is realized, a navigation scheme which adapts to a real-time environment and is better in comprehensive cost can be generated, and the efficiency and economy of inland river transportation scheduling are improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent scheduling technology for inland waterway transportation, specifically an inland waterway transportation management system based on AI-based voyage planning. Background Technology

[0002] In current inland waterway freight transport management, there is a widespread problem of information asymmetry between cargo availability and transport capacity. Existing information platforms mainly publish and display cargo and vessel information, and the matching function is mostly based on a few static rules such as vessel tonnage and position for filtering. This matching method is disconnected from the subsequent actual navigation task execution. It fails to fully consider key factors such as route feasibility and navigation costs during matching, resulting in the generated matching schemes often needing adjustment or even being unenforceable in subsequent actual scheduling, leading to low overall scheduling efficiency.

[0003] Traditional voyage planning relies heavily on the captain's personal experience or simple electronic navigation charts for route finding. These methods fail to systematically incorporate real-time environmental data such as water levels, current speeds, and weather conditions. The resulting plans are often static and idealized routes, unable to achieve a dynamic optimal balance between travel time, fuel consumption, and navigational safety in actual navigation, and lacking the ability to respond to sudden hydrological and meteorological conditions or temporary channel closures.

[0004] A technical solution is needed that can organically coordinate cargo ship matching with voyage depth planning, and accurately handle multi-source, dynamic waterway and environmental constraints in the planning process, thereby outputting a truly executable and cost-optimized transportation mission solution. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art; To this end, the present invention proposes an inland waterway transport management system based on AI-based route planning, comprising: The information aggregation module is used to receive and store cargo demand information uploaded by cargo owner clients and ship status information uploaded by ship owner clients. The cargo demand information includes cargo type, origin and destination, time requirements and cargo volume, and the ship status information includes ship position, deadweight tonnage, seaworthiness status and available time. The intelligent matching module is used to obtain the cargo demand information and vessel status information from the information aggregation center, and perform correlation analysis on the cargo and vessels according to the preset matching rules to generate an initial set of matching schemes. The manual review module is used to receive the review operation instructions from the dispatcher for the initial matching scheme set, and generate a confirmed transportation task according to the review operation instructions; The AI-powered route planning module is used to perform route planning calculations based on the confirmed transportation task, by calling inland waterway data and real-time environmental data, and to generate a detailed route plan that includes the navigation path, expected time nodes, and energy consumption predictions.

[0006] Preferably, the intelligent matching module includes: The rule configuration unit is used to store the preset matching rules, which include load tonnage threshold matching rules, time window compatibility judgment rules, and cargo suitability assessment rules. The information extraction unit is used to extract cargo type, origin and destination, time requirements and cargo volume from the cargo demand information, and to extract ship position, deadweight tonnage, seaworthiness status and available time from the ship status information. The matching calculation unit is used to input the cargo type, origin and destination, time requirements and cargo volume, along with the ship's position, deadweight tonnage, seaworthiness status and available time, into the preset matching rules for calculation, and output a preliminary list of matched ships and a matching score. The scheme generation unit is used to sort the list of preliminarily matched ships according to the matching score, select ships with matching scores higher than a set threshold to generate the initial matching scheme set, the initial matching scheme set including recommended ship identifiers, expected cargo load rates and time matching scores.

[0007] Preferably, the matching operation unit includes: The load determination subunit is used to compare the cargo volume with the deadweight tonnage, calculate the ship's load margin, and determine the load suitability according to the deadweight tonnage threshold matching rule. The time judgment subunit is used to compare the time requirement with the available time, calculate the time overlap interval, and judge the degree of satisfaction of the time window according to the time window compatibility judgment rule. The cargo type determination subunit is used to query the preset cargo suitable vessel type library according to the cargo type, compare it with the current vessel type, and determine the suitability level between the cargo and the vessel according to the cargo suitability assessment rules. The comprehensive scoring subunit is used to perform weighted fusion calculations on the load adaptability, the degree of satisfaction of the time window, and the compatibility level between cargo and ship, and generate the matching score for each ship.

[0008] Preferably, the AI ​​flight planning module includes: The data acquisition unit is used to acquire the origin and destination information of the confirmed transportation task, and simultaneously acquire waterway hydrological data, lock distribution data and bridge clearance data from the inland waterway database, as well as acquire real-time hydrological and meteorological data from the external environment interface. The path planning unit is used to calculate multiple candidate navigation paths based on the origin and destination information, waterway hydrological data, lock distribution data, bridge clearance data, and real-time hydrological and meteorological data, using a preset path search algorithm. The evaluation and optimization unit is used to calculate the comprehensive cost of each candidate navigation path according to a preset evaluation model. The comprehensive cost includes navigation time cost, energy consumption cost and risk cost, and selects the candidate navigation path with the lowest comprehensive cost as the preferred navigation path. The planning generation unit is used to calculate the expected arrival time of key waypoints and the predicted energy consumption for the entire voyage based on the preferred navigation route and the ship performance model, and integrate them to generate the detailed voyage plan.

[0009] Preferably, the evaluation and optimization unit includes: The time cost calculation subunit is used to calculate the navigation time cost based on the length of the preferred navigation route, the water flow velocity in the waterway hydrological data, the expected lock waiting time in the lock distribution data, and the still water speed in the ship performance model. The energy consumption cost calculation subunit is used to calculate the energy consumption cost based on the length of the preferred navigation path, the water flow speed and direction in the waterway hydrological data, the main engine fuel consumption characteristic curve in the ship performance model, and the wind direction and wind speed in the real-time hydrological and meteorological data. The risk cost calculation subunit is used to assess navigation risks and quantify the risk costs based on the relationship between the bridge clearance data and the ship's draft and height, the distribution information of shoals and dangerous shoals in the waterway hydrological data, and the visibility and wind force level in the real-time hydrological and meteorological data. The cost fusion subunit is used to linearly combine the travel time cost, energy consumption cost, and risk cost according to preset weights to calculate the comprehensive cost of each candidate travel path.

[0010] Preferably, the system further includes: The mission execution monitoring module is used to receive real-time location information reported by the ship positioning device during the execution of the transportation mission, compare the real-time location information with the expected time nodes in the detailed voyage plan, and generate a navigation status deviation report. The dynamic replanning trigger module is used to send a replanning request to the AI ​​voyage planning engine when the deviation value in the navigation status deviation report exceeds a preset tolerance threshold. The replanning request includes the current ship position and the remaining mission path.

[0011] Preferably, the task execution monitoring module includes: The data receiving subunit is used to continuously receive the real-time location information and ship status report from the ship positioning device; The comparison and analysis subunit is used to extract the expected time node and expected position corresponding to the current flight segment from the detailed flight plan, compare them with the real-time position information and the current system time, and calculate the time deviation value and position deviation distance. The report generation subunit is used to generate a navigation status deviation report, which includes the deviation level, the speculation of the deviation cause, and the impact assessment, based on the time deviation value and the position deviation distance, combined with the preset navigation status assessment rules.

[0012] Preferably, the dynamic replanning triggering module includes: The threshold judgment subunit is used to read the time deviation value and position deviation distance in the navigation status deviation report and compare them with the preset tolerance thresholds for time and distance, respectively. A request generation subunit is used to generate the replanning request when either the time deviation value or the position deviation distance exceeds its corresponding tolerance threshold. The replanning request encapsulates the current ship position, mission destination, remaining cargo information, and the latest real-time environmental data request instruction. The instruction sending subunit is used to send the replanning request to the AI ​​flight planning engine and trigger the AI ​​flight planning engine to start a new round of flight planning calculation based on the current state.

[0013] Preferably, the system further includes: The data feedback and learning module is used to collect detailed flight plans, actual flight trajectory data, actual energy consumption data, and flight status deviation reports for completed transportation tasks. The model optimization unit is used to compare and analyze the predicted data in the detailed voyage plan with the actual voyage trajectory data and actual energy consumption data, calculate the prediction error, and use the prediction error to iteratively optimize the parameters of the ship performance model and evaluation model used in the AI ​​voyage planning engine.

[0014] Preferably, the model optimization unit includes: The error calculation subunit is used to extract the expected time nodes and energy consumption prediction values ​​from the detailed flight plan for each completed transportation task, and compare them point by point with the actual arrival time nodes extracted from the actual flight trajectory data and the actual energy consumption values ​​extracted from the actual energy consumption data to generate a time prediction error sequence and an energy consumption prediction error sequence. The parameter adjustment subunit is used to adjust the speed-resistance relationship parameters in the ship performance model and the calculation weights of time cost and energy cost in the evaluation model based on the time prediction error sequence and energy consumption prediction error sequence using the gradient descent algorithm. The model update subunit is used to update the adjusted parameters into the model corresponding to the AI ​​flight planning engine for flight planning calculations of subsequent transportation tasks.

[0015] Compared with the prior art, the beneficial effects of the present invention are: The intelligent matching module incorporates rapid voyage assessment logic based on waterway data when performing cargo and vessel correlation analysis. While applying basic matching rules such as load and time, this module performs preliminary simulation calculations of navigation costs and feasibility for candidate transport routes. This ensures that the generated initial matching scheme set has undergone a round of voyage-level screening, eliminating obviously infeasible or costly combinations. Dispatchers are then faced with higher-quality pre-selected schemes that are closer to the final executable state, reducing manual decision-making time and creating a positive feedback loop and continuous optimization in the system's scheme output from matching to planning.

[0016] The AI-powered route planning module establishes a route planning model that deeply integrates static inland waterway attribute data with dynamic real-time environmental monitoring data. The model quantifies information into specific impact coefficients on ship navigation resistance, safe passage, and energy consumption, constructing a dynamic optimization function with comprehensive navigation cost as the objective and waterway conditions and safety rules as constraints. By solving this function, a detailed route and timetable can be generated that achieves optimal overall navigation efficiency and fuel economy under current and predicted conditions, while ensuring safety. This plan is no longer a fixed route but an adaptive scheme that can be dynamically adjusted based on environmental inputs. Attached Figure Description

[0017] Figure 1 This is a timeline diagram of the inland waterway transport management system based on AI-driven route planning as described in this invention. Figure 2 A flowchart illustrating how the intelligent matching module works; Figure 3 A flowchart illustrating the work done for the AI ​​flight planning module; Figure 4 A two-bar chart showing the risk level and tolerance threshold of a flight segment; Figure 5 A bar chart comparing the predicted and actual energy consumption of transportation tasks. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] See Figure 1 The specific implementation of the inland waterway transportation management system based on AI-powered route planning is as follows: The information aggregation module receives and stores cargo demand information uploaded by cargo owners and vessel status information uploaded by ship owners. The cargo demand information includes cargo type, origin and destination, time requirements, and cargo volume. The vessel status information includes vessel location, deadweight tonnage, seaworthiness, and available time. The intelligent matching module obtains cargo demand information and vessel status information from the information aggregation center and performs correlation analysis on cargo and vessels according to preset matching rules to generate an initial matching scheme set. The manual review module receives the dispatcher's review operation instructions for the initial matching scheme set and generates a confirmed transportation task according to the review operation instructions. The AI-powered route planning module calls inland waterway data and real-time environmental data based on the confirmed transportation task to perform route planning calculations and generate a detailed route plan including navigation path, expected time nodes, and energy consumption prediction.

[0020] In one embodiment of the present invention, see [reference] Figure 2 The intelligent matching module begins operation. The information extraction unit extracts fields such as cargo type, origin and destination, time requirements, and cargo volume from the cargo demand information provided by the information aggregation module, and extracts fields such as ship position, deadweight tonnage, seaworthiness status, and available time from the ship status information. The rule configuration unit stores and maintains preset matching rules, including deadweight tonnage threshold matching rules for judging ship carrying capacity, time window compatibility judgment rules for judging the degree of schedule matching, and cargo suitability assessment rules for judging cargo transportation safety requirements. In some embodiments, the deadweight tonnage threshold matching rule defines an allowable load margin range, the time window compatibility judgment rule defines a minimum proportion threshold for time overlap, and the cargo suitability assessment rule is associated with a preset cargo suitable vessel type library, which defines the ship types or classes allowed to be transported for each type of cargo.

[0021] The matching and processing unit receives the extracted fields from the information extraction unit and performs calculations. The load factor judgment subunit compares the cargo volume in the cargo demand information with the deadweight tonnage in the ship status information, calculates the ship's load margin, judges the load factor suitability according to the deadweight tonnage threshold matching rules, and generates a quantitative load factor suitability score. The time judgment subunit compares the time requirement in the cargo demand information with the available time in the ship status information, calculates the length of the time overlap interval between the two, calculates the degree of satisfaction of the time window according to the time window compatibility judgment rules, and generates a time window satisfaction score. The cargo type judgment subunit queries the preset cargo suitable ship type library according to the cargo type in the cargo demand information, obtains the list of ship types allowed for that cargo type, compares the currently processed ship type with this list, judges the suitability level between the cargo and the ship according to the cargo suitability assessment rules, and generates a cargo suitability score.

[0022] The comprehensive scoring subunit performs a weighted fusion calculation on the load adaptability score, time window satisfaction score, and cargo adaptability score to generate a matching score for each vessel. It can be understood that the weighting coefficients used in the weighted fusion calculation can be preset in the rule configuration unit. In some embodiments, the matching score is calculated using the following formula: in: Represents the match score. Represents load-bearing adaptability score. The score represents the degree of satisfaction within the time window. Represents the goods' suitability score. , , It is a preset weighting coefficient and satisfies Optionally, load adaptability score. The calculations take into account whether the ship's deadweight margin is within a preset reasonable range, and cargo suitability scoring. The system assigns different scores based on the matching results between the vessel type and the cargo's suitable vessel type database. The scheme generation unit receives the preliminary matched vessel list and matching score for each vessel from the matching calculation unit. It sorts the preliminary matched vessel list in descending order based on the matching score, selects vessel records with matching scores higher than a set threshold, and generates an initial matching scheme set for each cargo demand. This set includes fields such as recommended vessel identifier, expected load factor, and time matching degree. The expected load factor is calculated as the ratio of cargo volume to vessel deadweight tonnage, and the time matching degree is calculated as the ratio of the length of the time overlap interval to the total length of the time requirements in the cargo demand information. Optionally, the initial matching scheme set may provide multiple vessel options with high matching degrees for each cargo demand for selection by the subsequent manual review module.

[0023] In one embodiment of the present invention, see [reference] Figure 3 In practical implementation, the AI ​​route planning module receives confirmed transportation tasks from the manual review module. The data acquisition unit first parses the confirmed transportation tasks to obtain origin and destination information, which includes the specific origin and destination port terminals. Simultaneously, the data acquisition unit retrieves waterway hydrological data, lock distribution data, and bridge clearance data related to the origin and destination from the inland waterway database. Waterway hydrological data includes static water depth, average current velocity, and current direction for different sections. Lock distribution data includes lock locations and average waiting times for passage. Bridge clearance data includes bridge locations and navigable clearance height. The data acquisition unit also obtains real-time hydrological and meteorological data covering the planned navigation area from an external environment interface. This real-time hydrological and meteorological data includes real-time water flow velocity, wind direction and speed, and visibility. In some embodiments, the inland waterway database is stored in the form of electronic waterway charts, and the external environment interface is connected to real-time data services provided by maritime or meteorological departments.

[0024] The route planning unit performs calculations based on the origin and destination information, waterway hydrological data, lock distribution data, bridge clearance data, and real-time hydrological and meteorological data provided by the data acquisition unit. Using a pre-defined route search algorithm, the unit searches the waterway network model for all feasible routes from the origin to the destination, generating multiple candidate navigation paths. The pre-defined route search algorithm is the Dijkstra algorithm, and each candidate navigation path consists of a series of continuous waterway segments and a sequence of necessary lock nodes. The evaluation and optimization unit evaluates each candidate navigation path output by the route planning unit, calculating the comprehensive cost of each candidate navigation path according to a pre-defined evaluation model. The comprehensive cost consists of three parts: navigation time cost, energy consumption cost, and risk cost. The evaluation and optimization unit selects the candidate navigation path with the lowest comprehensive cost as the preferred navigation path for this planning process.

[0025] In practical implementation, the time cost calculation subunit within the evaluation and optimization unit is responsible for calculating the navigation time cost. Based on the total length of the optimized navigation path, the current velocity of each segment in the waterway hydrological data, the estimated waiting time for each lock node in the lock distribution data, and the ship's speed performance curve under still water conditions stored in the ship performance model, this subunit calculates the estimated total navigation time required to complete the entire path and converts the estimated total navigation time into a time cost value. The energy consumption cost calculation subunit is responsible for calculating the energy consumption cost. Based on the same optimized navigation path length, combined with detailed current velocity and direction data from the waterway hydrological data, the relationship curves between main engine fuel consumption, speed, and load stored in the ship performance model, and overlaying the influence model of wind direction and speed on ship resistance from real-time hydrological and meteorological data, this subunit calculates the estimated total fuel consumption to complete the transportation task and converts the estimated total fuel consumption into an energy consumption cost value.

[0026] The risk cost calculation subunit is responsible for quantifying the navigation risk cost. It reads the bridge clearance data and compares it with the ship's height above water and draft in the ship status information of the vessel designated for the transportation task. It identifies bridge sections with risks of insufficient navigation clearance or excessive draft. The risk cost calculation subunit also reads the distribution information of shoals and dangerous shoals marked in the waterway hydrological data, and integrates the visibility level and wind force data released in real-time hydrological and meteorological data to assess the probability and severity of risks such as grounding, collision, and severe weather that may be encountered during navigation. Finally, it outputs a comprehensive risk quantification value as the risk cost.

[0027] In some embodiments, the risk cost calculation assigns different risk coefficients to different types of risk events. The cost fusion subunit receives the time cost, energy cost, and risk cost values ​​output by the time cost calculation subunit, energy consumption cost calculation subunit, and risk cost calculation subunit, respectively. The cost fusion subunit linearly combines these three values ​​according to a preset weight allocation to calculate the final comprehensive cost value for each candidate flight path. Optionally, the comprehensive cost value... The calculation formula is: in: Represents the overall cost, Represents the cost of travel time. Represents the cost of energy consumption. Represents risk and cost. , , It is a preset weighting coefficient and satisfies The planning generation unit calculates the expected arrival time sequence of key waypoints (such as important turning points, locks, and bridges) during the voyage, based on the selected preferred navigation path and parameters from the ship's performance model regarding acceleration, deceleration, and cornering performance. Simultaneously, it generates a predicted energy consumption value for the entire journey based on intermediate calculation results from the energy consumption cost calculation subunit. The planning generation unit integrates the detailed description of the preferred navigation path, the expected time sequence, and the predicted energy consumption value to generate a complete and detailed voyage plan. This detailed voyage plan will be output in a structured data format. Optionally, the detailed voyage plan can also be visualized on an electronic nautical chart.

[0028] In one embodiment of the present invention, the system activates the task execution monitoring module during the transportation task execution phase. The task execution monitoring module continuously receives real-time location information and ship status reports from the ship positioning device via a data receiving subunit. The ship status report includes the ship's current speed, heading, and main engine operating status. The data receiving subunit parses and formats the received real-time location information and ship status reports. A comparison and analysis subunit retrieves the detailed voyage plan bound to the transportation task from the system storage. The detailed voyage plan includes a series of preset estimated time nodes and corresponding estimated position coordinate sequences. The comparison and analysis subunit determines the planned voyage segment the ship should be on based on the current system time and extracts the estimated time node and estimated position corresponding to that segment. The comparison and analysis subunit compares the real-time location information reported by the ship positioning device with the estimated position to calculate the position deviation distance, and simultaneously compares the current system time with the estimated time node to calculate the time deviation value.

[0029] The report generation subunit receives the time deviation value and position deviation distance calculated by the comparison and analysis subunit. Based on preset navigation status assessment rules, the subunit analyzes and classifies the deviations. These rules define time deviation thresholds and position deviation distance thresholds for different levels of deviation. In some embodiments, the navigation status assessment rules classify deviation levels into "normal," "concern," and "warning" levels. Based on the time deviation value, position deviation distance, and any abnormal information in potential ship status reports (such as main engine shutdown reports), the report generation subunit logically infers the cause of the deviation. Finally, the subunit generates a structured navigation status deviation report, which includes the currently identified deviation level, the inferred cause of the deviation, and an assessment of its potential impact on subsequent planned voyages. It is understood that the inferred cause of the deviation may include "water flow speed higher than predicted," "ship mechanical failure," or "excessive queuing time at the lock," etc.

[0030] In implementation, the dynamic replanning trigger module continuously monitors the output of the flight status deviation report. The threshold judgment subunit reads two core indicators from the latest flight status deviation report: the time deviation value and the position deviation distance. The threshold judgment subunit accesses a preset tolerance threshold configuration, which includes a time tolerance threshold and a distance tolerance threshold. The threshold judgment subunit compares the time deviation value with the time tolerance threshold and simultaneously compares the position deviation distance with the distance tolerance threshold. When either the time deviation value or the position deviation distance exceeds its corresponding tolerance threshold, the threshold judgment subunit determines that a deviation event requiring flight replanning has occurred. In some embodiments, the time tolerance threshold and the distance tolerance threshold can be set differently according to the risk level of different flight segments.

[0031] The request generation subunit begins operation upon receiving a trigger signal from the threshold judgment subunit. It acquires the ship's current precise position, the initial destination of the transport mission, and information on remaining cargo, and generates a request instruction for the latest real-time environmental data to the external environment interface. The request generation subunit encapsulates this information into a structured replanning request data packet. Optionally, the replanning request data packet also includes the specific deviation type and value that triggered the trigger. The instruction sending subunit is responsible for sending the replanning request data packet to the AI ​​route planning engine within the AI ​​route planning module. The sending action of the instruction sending subunit simultaneously constitutes a trigger signal. Upon receiving the replanning request data packet, the AI ​​route planning engine immediately initiates a new round of route planning calculations based on the current ship position encapsulated in the data packet as the new starting point and the original destination as the endpoint, combined with the latest real-time environmental data obtained from the request.

[0032] In one embodiment of the present invention, the threshold judgment subunit of the dynamic replanning triggering module directly reads and parses the navigation status deviation report generated by the task execution monitoring module. The navigation status deviation report contains two quantitative indicators: the calculated time deviation value and the position deviation distance. The threshold judgment subunit internally maintains or accesses a preset tolerance threshold configuration table. This configuration table explicitly defines the condition thresholds that must be met to trigger replanning. The threshold judgment subunit compares the time deviation value in the navigation status deviation report with the time tolerance threshold defined in the configuration table, and simultaneously compares the position deviation distance in the navigation status deviation report with the distance tolerance threshold defined in the configuration table. The comparison operations are parallel logical judgments. In some embodiments, the tolerance threshold configuration table can be set differently according to the geographical characteristics or risk level of the flight segment. For example, a stricter tolerance threshold can be used in high-risk narrow channels. Referring to Table 1, a possible tolerance threshold configuration table is shown.

[0033] Table 1: Correspondence between Flight Segment Risk Levels and Tolerance Thresholds Leg risk level Time tolerance threshold (minutes) Distance tolerance threshold (km) Low risk 60 5 Medium risk 30 2 High risk 15 1 The request generation subunit monitors the output status of the threshold judgment subunit. When the threshold judgment subunit's comparison logic determines that either the time deviation exceeds the time tolerance threshold or the position deviation exceeds the distance tolerance threshold, the threshold judgment subunit sends a trigger signal to the request generation subunit, which then initiates the replanning request construction process. The request generation subunit first obtains the ship's latest precise latitude and longitude coordinates from the system cache or positioning data stream as the ship's current position, extracts the mission destination information from the original transportation task order, and obtains remaining cargo information, such as remaining cargo quantity, from the ship status report. Simultaneously, the request generation subunit generates an instruction to request the data acquisition unit to immediately pull the latest real-time hydrological and meteorological data from the external environment interface. This real-time hydrological and meteorological data is crucial for the subsequent accurate replanning by the AI ​​journey planning engine. The request generation subunit encapsulates all the above information into a structured data packet, which constitutes the replanning request.

[0034] In specific implementation, the instruction sending subunit is responsible for transmitting and triggering replanning requests. The instruction sending subunit receives the replanning request data packet encapsulated by the request generation subunit. Through a message queue or service interface defined within the system, the instruction sending subunit sends the replanning request data packet to the core processing unit of the AI ​​flight path planning module, namely the AI ​​flight path planning engine. The sending operation itself is accompanied by a clear calling instruction, which instructs the AI ​​flight path planning engine to immediately interrupt any background computation tasks it may be performing and switch to processing the newly received replanning request. Optionally, the dynamic replanning triggering module can be designed to allow multiple triggers within a single transport mission. That is, after each replanning request is generated and sent, the threshold judgment subunit resets its judgment state and continues to perform a new round of monitoring and judgment based on subsequently updated flight status deviation reports. It is understood that, in order to assess the overall trend of the deviation or prevent false triggers caused by instantaneous fluctuations, in some embodiments, the judgment logic of the threshold judgment subunit can introduce a continuous monitoring window. For example, it requires that the deviation value exceeds the tolerance threshold for N consecutive monitoring periods to be considered a valid trigger. The judgment formula can be expressed as: in: Represents the final trigger judgment result (Boolean value). This represents the time deviation value for the i-th monitoring period. This represents the position deviation distance during the i-th monitoring period. Represents the time tolerance threshold. This represents the distance tolerance threshold. Represents the total number of monitoring periods. The minimum number of cycles required to satisfy the trigger condition within the window is represented by the symbol. Represents the logical "OR" operation. The instruction sending subunit only... The operation of sending a replanning request will only be executed if the result is true.

[0035] See Figure 4 This is a dual-bar chart showing the relationship between waterway segment risk level and tolerance threshold. As the risk level of a waterway segment increases, both the time and distance tolerance thresholds decrease, reflecting the dynamic replanning triggering logic of "higher risk, stricter monitoring." This chart visually illustrates the "risk-threshold" matching rules in the inland waterway transport management system, helping dispatchers quickly understand the deviation tolerance standards for different waterways and providing visualized parameter data for the system's automatic replanning triggering. By transforming the "correlation logic between waterway segment risk and tolerance threshold" from abstract text into an intuitive chart, it clearly presents the core rule of "higher risk, stricter threshold," reducing the cost of understanding the system's triggering mechanism.

[0036] In one embodiment of the invention, the data feedback and learning module is activated upon completion of each transportation task. This module systematically collects all process and result data related to the task from the system database. This data includes a detailed voyage plan generated by the AI ​​voyage planning module at the start of the task, actual navigation trajectory data transmitted back by the ship's positioning device, actual energy consumption data recorded by the ship's fuel monitoring system, and navigation status deviation reports generated by the task execution monitoring module. The data feedback and learning module cleans, aligns, and timestamps the collected multi-source heterogeneous data, forming a complete data record package indexed by a single transportation task. In some embodiments, the data record package is stored in the form of a structured database table or file for easy subsequent batch analysis and processing.

[0037] The model optimization unit loads the data record package organized by the data feedback and learning module. The error calculation subunit performs error analysis on the data record package for each completed transportation task. The error calculation subunit extracts the expected arrival time sequence for each key waypoint and the predicted energy consumption value for the entire journey from the detailed voyage plan. Simultaneously, it parses the actual arrival timestamp sequence of the ship at each key waypoint from the actual navigation trajectory data and extracts the total actual fuel consumption for the entire mission from the actual energy consumption data. The error calculation subunit compares the expected arrival time with the actual arrival timestamp point by point, calculating the time prediction error for each waypoint and forming a time prediction error sequence; it also compares the predicted energy consumption value with the actual energy consumption value to calculate the energy consumption prediction error. Optionally, the error calculation subunit also calculates derived indicators such as the overall mission completion time error and the average speed prediction error.

[0038] The parameter adjustment subunit receives the time prediction error sequence and energy consumption prediction error sequence output by the error calculation subunit. Aiming to reduce these prediction errors, the subunit iteratively optimizes the parameters of the mathematical model used within the AI ​​journey planning engine. It can be understood that the core objects of optimization are the parameters in the ship performance model that describe the relationship between ship speed, resistance, and fuel consumption, as well as the weighting coefficients used in the evaluation model to calculate time and energy costs. In some embodiments, the parameter adjustment subunit uses the gradient descent algorithm as the optimization method, updating the parameter values ​​along the reverse direction of the gradient by calculating the gradient of the prediction error relative to the model parameters. The parameter update rule executed by the parameter adjustment subunit can be described as follows: in: This represents the updated model parameter vector. This represents the model parameter vector before the update. This represents the learning rate, i.e., the step size for parameter updates. Represents the loss function Regarding model parameters gradient, loss function It is jointly defined by time prediction error and energy consumption prediction error. The parameter adjustment subunit uses the gradient calculated from the current transportation task data to perform an iterative update on the relevant parameters.

[0039] The model update subunit is responsible for integrating the new parameters optimized by the parameter tuning subunit into the online system. The model update subunit obtains the updated ship performance model parameter set and evaluation model weight coefficient set from the parameter tuning subunit. It then securely writes these new parameters into the AI ​​voyage planning engine's model parameter configuration library, overwriting the old parameter version. This update is effective immediately; subsequent newly initiated transportation tasks will automatically use the optimized new parameters when calling the AI ​​voyage planning engine for voyage planning calculations.

[0040] See Figure 5 This is a bar chart comparing the predicted and actual energy consumption of a transportation task. It shows the energy consumption fluctuations of different transportation tasks. By combining information such as the task's route and cargo type, common characteristics of high-energy-consuming tasks can be analyzed, assisting in optimizing route planning to reduce actual energy costs. If the deviation between the prediction and actual energy consumption is within a reasonable range, the energy consumption prediction function of the AI ​​flight planning module can be verified as having practical value. If the deviation is too large, the data feedback and learning module needs to be triggered to optimize the model parameters. By identifying the difference between the predicted and actual energy consumption, the deviation scenarios of the model can be accurately located, providing clear optimization targets for the data feedback and learning module, and promoting the iterative upgrade of the energy consumption prediction model.

[0041] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An inland waterway transport management system based on AI-driven route planning, characterized in that: include: The information aggregation module is used to receive and store cargo demand information uploaded by cargo owner clients and ship status information uploaded by ship owner clients. The cargo demand information includes cargo type, origin and destination, time requirements and cargo volume, and the ship status information includes ship position, deadweight tonnage, seaworthiness status and available time. The intelligent matching module is used to obtain the cargo demand information and vessel status information from the information aggregation center, and perform correlation analysis on the cargo and vessels according to the preset matching rules to generate an initial set of matching schemes. The manual review module is used to receive the review operation instructions from the dispatcher for the initial matching scheme set, and generate a confirmed transportation task according to the review operation instructions; The AI-powered route planning module is used to perform route planning calculations based on the confirmed transportation task, by calling inland waterway data and real-time environmental data, and to generate a detailed route plan that includes the navigation path, expected time nodes, and energy consumption predictions.

2. The inland waterway transport management system based on AI-driven route planning according to claim 1, characterized in that, The intelligent matching module includes: The rule configuration unit is used to store the preset matching rules, which include load tonnage threshold matching rules, time window compatibility judgment rules, and cargo suitability assessment rules. The information extraction unit is used to extract cargo type, origin and destination, time requirements and cargo volume from the cargo demand information, and to extract ship position, deadweight tonnage, seaworthiness status and available time from the ship status information. The matching calculation unit is used to input the cargo type, origin and destination, time requirements and cargo volume, along with the ship's position, deadweight tonnage, seaworthiness status and available time, into the preset matching rules for calculation, and output a preliminary list of matched ships and a matching score. The scheme generation unit is used to sort the list of preliminarily matched ships according to the matching score, select ships with matching scores higher than a set threshold to generate the initial matching scheme set, the initial matching scheme set including recommended ship identifiers, expected cargo load rates and time matching scores.

3. The inland waterway transport management system based on AI-driven route planning according to claim 2, characterized in that, The matching operation unit includes: The load determination subunit is used to compare the cargo volume with the deadweight tonnage, calculate the ship's load margin, and determine the load suitability according to the deadweight tonnage threshold matching rule. The time judgment subunit is used to compare the time requirement with the available time, calculate the time overlap interval, and judge the degree of satisfaction of the time window according to the time window compatibility judgment rule. The cargo type determination subunit is used to query the preset cargo suitable vessel type library according to the cargo type, compare it with the current vessel type, and determine the suitability level between the cargo and the vessel according to the cargo suitability assessment rules. The comprehensive scoring subunit is used to perform weighted fusion calculations on the load adaptability, the degree of satisfaction of the time window, and the compatibility level between cargo and ship, and generate the matching score for each ship.

4. The inland waterway transport management system based on AI-driven route planning according to claim 1, characterized in that, The AI ​​flight planning module includes: The data acquisition unit is used to acquire the origin and destination information of the confirmed transportation task, and simultaneously acquire waterway hydrological data, lock distribution data and bridge clearance data from the inland waterway database, as well as acquire real-time hydrological and meteorological data from the external environment interface. The path planning unit is used to calculate multiple candidate navigation paths based on the origin and destination information, waterway hydrological data, lock distribution data, bridge clearance data, and real-time hydrological and meteorological data, using a preset path search algorithm. The evaluation and optimization unit is used to calculate the comprehensive cost of each candidate navigation path according to a preset evaluation model. The comprehensive cost includes navigation time cost, energy consumption cost and risk cost, and selects the candidate navigation path with the lowest comprehensive cost as the preferred navigation path. The planning generation unit is used to calculate the expected arrival time of key waypoints and the predicted energy consumption for the entire voyage based on the preferred navigation route and the ship performance model, and to integrate and generate the detailed voyage plan.

5. The inland waterway transport management system based on AI-driven route planning according to claim 4, characterized in that, The evaluation and optimization unit includes: The time cost calculation subunit is used to calculate the navigation time cost based on the length of the preferred navigation route, the water flow velocity in the waterway hydrological data, the expected lock waiting time in the lock distribution data, and the still water speed in the ship performance model. The energy consumption cost calculation subunit is used to calculate the energy consumption cost based on the length of the preferred navigation path, the water flow speed and direction in the waterway hydrological data, the main engine fuel consumption characteristic curve in the ship performance model, and the wind direction and wind speed in the real-time hydrological and meteorological data. The risk cost calculation subunit is used to assess navigation risks and quantify the risk costs based on the relationship between the bridge clearance data and the ship's draft and height, the distribution information of shoals and dangerous shoals in the waterway hydrological data, and the visibility and wind force level in the real-time hydrological and meteorological data. The cost fusion subunit is used to linearly combine the travel time cost, energy consumption cost, and risk cost according to preset weights to calculate the comprehensive cost of each candidate travel path.

6. The inland waterway transport management system based on AI-driven route planning according to claim 1, characterized in that, The system also includes: The mission execution monitoring module is used to receive real-time location information reported by the ship positioning device during the execution of the transportation mission, compare the real-time location information with the expected time nodes in the detailed voyage plan, and generate a navigation status deviation report. The dynamic replanning trigger module is used to send a replanning request to the AI ​​voyage planning engine when the deviation value in the navigation status deviation report exceeds a preset tolerance threshold. The replanning request includes the current ship position and the remaining mission path.

7. The inland waterway transport management system based on AI-driven route planning according to claim 6, characterized in that, The task execution monitoring module includes: The data receiving subunit is used to continuously receive the real-time location information and ship status report from the ship positioning device; The comparison and analysis subunit is used to extract the expected time node and expected position corresponding to the current flight segment from the detailed flight plan, compare them with the real-time position information and the current system time, and calculate the time deviation value and position deviation distance. The report generation subunit is used to generate a navigation status deviation report, which includes the deviation level, the speculation of the deviation cause, and the impact assessment, based on the time deviation value and the position deviation distance, combined with the preset navigation status assessment rules.

8. The inland waterway transport management system based on AI-driven route planning according to claim 7, characterized in that, The dynamic replanning trigger module includes: The threshold judgment subunit is used to read the time deviation value and position deviation distance in the navigation status deviation report and compare them with the preset tolerance thresholds for time and distance, respectively. A request generation subunit is used to generate the replanning request when either the time deviation value or the position deviation distance exceeds its corresponding tolerance threshold. The replanning request encapsulates the current ship position, mission destination, remaining cargo information, and the latest real-time environmental data request instruction. The instruction sending subunit is used to send the replanning request to the AI ​​flight planning engine and trigger the AI ​​flight planning engine to start a new round of flight planning calculation based on the current state.

9. The inland waterway transport management system based on AI-driven route planning according to claim 1, characterized in that, The system also includes: The data feedback and learning module is used to collect detailed flight plans, actual flight trajectory data, actual energy consumption data, and flight status deviation reports for completed transportation tasks. The model optimization unit is used to compare and analyze the predicted data in the detailed voyage plan with the actual voyage trajectory data and actual energy consumption data, calculate the prediction error, and use the prediction error to iteratively optimize the parameters of the ship performance model and evaluation model used in the AI ​​voyage planning engine.

10. The inland waterway transport management system based on AI-driven route planning according to claim 9, characterized in that, The model optimization unit includes: The error calculation subunit is used to extract the expected time nodes and energy consumption prediction values ​​from the detailed flight plan for each completed transportation task, and compare them point by point with the actual arrival time nodes extracted from the actual flight trajectory data and the actual energy consumption values ​​extracted from the actual energy consumption data to generate a time prediction error sequence and an energy consumption prediction error sequence. The parameter adjustment subunit is used to adjust the speed-resistance relationship parameters in the ship performance model and the calculation weights of time cost and energy cost in the evaluation model based on the time prediction error sequence and energy consumption prediction error sequence using the gradient descent algorithm. The model update subunit is used to update the adjusted parameters into the model corresponding to the AI ​​flight planning engine for flight planning calculations of subsequent transportation tasks.

Images