Intelligent navigation and emergency decision-making method and system for complex channel ship

Abstract

The invention relates to the technical field of intelligent navigation and control of ships. The invention provides a complex channel ship intelligent navigation and emergency decision-making method and system. The method comprises the following steps: acquiring environment data through a multi-source heterogeneous sensor array, and establishing a channel three-dimensional dynamic environment model; establishing a multi-objective optimization function, and performing dynamic path planning by adopting an improved model prediction control algorithm; synchronizing motion state parameters of an actual ship and a virtual ship model in real time, constructing an emergency decision tree in combination with an expert knowledge base, and verifying the feasibility of an emergency decision through Monte Carlo simulation; carrying out local route optimization by adopting edge computing nodes, carrying out multi-ship trajectory prediction through a federated learning mechanism, and generating a corresponding collaborative collision avoidance strategy; and establishing a dynamic priority scheduling mechanism, implementing hierarchical response, and confirming a global avoidance scheme through a distributed consensus algorithm. The problems that an existing inland ship intelligent system is limited in perception, rigid in decision and weak in collaboration in a complex scene are solved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent navigation and control technology for ships, and in particular to a method and system for intelligent navigation and emergency decision-making for ships in complex waterways. Background Art

[0002] Inland waterways face severe challenges in navigation safety and efficiency due to their narrow, winding, densely packed, and complex hydrological characteristics. Existing ship intelligent systems have significant deficiencies in complex scenarios:

[0003] Perception limitations: reliance on a single sensor is susceptible to interference, multi-source data fusion has low accuracy, and environmental modeling is incomplete;

[0004] Rigid decision-making: Traditional path planning algorithms struggle to cope with dynamic avoidance and hydrological coupling constraints. Machine learning methods lack embedded ship physical models, and emergency response relies on manual experience.

[0005] Weak coordination: Ship-shore communication has high latency, the vessel traffic management system mainly relies on one-way instructions, and the global coordinated collision avoidance and resource scheduling capabilities are insufficient. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for intelligent navigation and emergency decision-making of ships in complex waterways, aiming to solve the problems of perception limitations, rigid decision-making and weak coordination of existing inland ship intelligent systems in complex scenarios.

[0007] The present invention is achieved through the following technical solutions:

[0008] A method for intelligent navigation and emergency decision-making of ships in complex waterways, comprising the following steps:

[0009] Environmental data is collected through a multi-source heterogeneous sensor array, and an adaptive Kalman filter algorithm is used to perform multi-sensor spatiotemporal calibration and data compensation to establish a three-dimensional dynamic environment model of the waterway.

[0010] Based on the ship's kinematic model and real-time hydrodynamic parameters, combined with the three-dimensional dynamic environment model of the channel, a multi-objective optimization function is established that includes ship maneuverability constraints, channel boundary constraints, and navigation rules. An improved model predictive control algorithm is used for dynamic path planning.

[0011] Based on dynamic path planning, the motion state parameters of the actual ship and the virtual ship model are synchronized in real time. An emergency decision tree is constructed in combination with the expert knowledge base, and the feasibility of emergency decision-making is verified through Monte Carlo simulation.

[0012] Edge computing nodes are used to perform local route optimization for dynamic path planning, and a federated learning mechanism is used to predict multi-ship trajectories and generate corresponding collaborative collision avoidance strategies.

[0013] Based on the collaborative collision avoidance strategy, a dynamic priority scheduling mechanism is established. When navigation conflicts or sudden hydrological changes are detected, a graded response is implemented based on the ship's cargo level, draft and power characteristics, and the global avoidance plan is confirmed through a distributed consensus algorithm.

[0014] Optionally, the specific process of collecting environmental data through a multi-source heterogeneous sensor array, performing multi-sensor spatiotemporal calibration and data compensation using an adaptive Kalman filter algorithm, and establishing a three-dimensional dynamic environment model of the waterway is as follows:

[0015] A multi-source heterogeneous sensor array consisting of radar, lidar, visual camera, water depth sensor and ship automatic identification system is arranged at the bow, stern and both sides of the ship to collect channel boundaries, obstacle locations, ship motion parameters, water depth information and the identity status data of navigable ships;

[0016] Perform time stamp synchronization and spatial coordinate system conversion on the output data of each sensor to establish a unified time and space benchmark;

[0017] An adaptive adjustment model is constructed that includes the sensor measurement noise covariance matrix and the system process noise covariance matrix. The matrix parameters are dynamically adjusted according to the real-time environmental interference intensity. The measurement data of different sensors are iteratively processed using Kalman filtering to achieve spatiotemporal calibration and abnormal data compensation.

[0018] The calibrated and compensated static channel boundary data, dynamic ship motion data and real-time hydrological data are gridded and voxelized to generate a three-dimensional dynamic environment model of the channel that includes channel water depth, three-dimensional coordinates of obstacles and dynamic trajectory of ships. The model is refreshed in real time through the data update interface.

[0019] Optionally, the specific process of constructing the adaptive adjustment model including the sensor measurement noise covariance matrix and the system process noise covariance matrix is:

[0020] Initialize the basic values of the sensor measurement noise covariance matrix and the system process noise covariance matrix based on the sensor's historical measurement data and the typical environmental parameters of the waterway;

[0021] The environmental interference monitoring module deployed in the sensor array collects multi-dimensional interference parameters including electromagnetic interference intensity, water turbidity and meteorological noise in real time, and constructs an interference intensity evaluation function.

[0022] A dynamic adjustment rule base for the noise covariance matrix is established. Based on the real-time interference intensity assessment results, the noise variance terms of each sensor's corresponding measurement dimension in the sensor measurement noise covariance matrix and the process noise terms of the ship kinematic model state transition in the system process noise covariance matrix are corrected through a fuzzy logic controller or an adaptive weight allocation algorithm.

[0023] The adjusted sensor measurement noise covariance matrix and system process noise covariance matrix are input into the Kalman filter iterative process to form a closed-loop feedback regulation mechanism to perform real-time adaptation to different sensor measurement noises and system dynamic noises.

[0024] Optionally, the specific process of establishing a multi-objective optimization function including ship maneuverability constraints, channel boundary constraints and navigation rules based on the ship kinematic model and real-time hydrodynamic parameters in combination with the three-dimensional dynamic environment model of the channel, and using the improved model predictive control algorithm for dynamic path planning is as follows:

[0025] The three-degree-of-freedom kinematic differential equation of the ship is established as shown in the following equation (1):

[0026]

[0027] Where (x, y) represents the position coordinates of the ship in the plane coordinate system; represents the velocity component of the ship along the x-axis; represents the velocity component of the ship along the y-axis; represents the ship's turning rate; ψ represents the ship's heading angle; u represents the ship's longitudinal speed along the stern direction of the hull; v represents the ship's transverse speed perpendicular to the stern direction of the hull; r represents the rate of change of the ship's heading angle; θ h represents the real-time hydrodynamic parameter set; Δ x (θ h ) represents the additional velocity component caused by the real-time hydrodynamic parameters in the x direction of the inertial coordinate system; Δ y (θ h ) represents the additional velocity component caused by the real-time hydrodynamic parameters in the y direction of the inertial coordinate system; Δ ψ (θ h ) represents the correction term for the rate of change of the ship's heading angle, which is used to compensate for the interference of hydrodynamic parameters on the ship's steering characteristics;

[0028] Construct a multi-objective optimization function as shown in the following formula (2):

[0029]

[0030] Among them, J represents the total cost function; N p Indicates the length of the prediction time domain; X k represents the ship state vector at step k; X ref,k represents the reference trajectory state vector of the kth step; Q represents the state tracking error weight matrix; U k represents the control input vector of the step; R represents the control input weight matrix; γ represents the obstacle repulsion coefficient; Ω k represents the state vector of the channel obstacle at step k; d(Xk ,Ω k ) represents the minimum Euclidean distance between the k-th step ship position and the channel obstacle; ρ represents the speed limit penalty factor; V lim Indicates the speed limit value of the channel; v k represents the actual speed of the ship at step k;

[0031] Define the time-varying constraint set as shown in the following formula (3):

[0032]

[0033] Where U(t) represents the control input variable of the ship at time t; U max (t) and U min (t) represents the maximum value vector and minimum value vector of the control input respectively; (x k ,y k ) represents the plane coordinate of the ship at the kth moment in the prediction time domain; represents the channel boundary constraint function; a gradient representing the channel geometry; Represents the position coordinate vector of the i-th obstacle; R safe represents the static safety radius; τ resp Indicates the braking response time;

[0034] The improved model predictive control algorithm is used to solve the constrained optimization function in the receding horizon, as shown in the following formula (4):

[0035]

[0036] Among them, f(X k ,U k ,θ h ) represents the state transfer function of the ship kinematic model, which is obtained by discretizing the differential equation of formula (1); X k+1 represents the ship state vector at step k+1;

[0037] Update θ through real-time hydrodynamic parameter identification module h The estimated value of is used to modify the ship kinematic model;

[0038] The augmented Lagrangian method is used to deal with the channel boundary constraints, as shown in the following formula (5):

[0039]

[0040] Among them, J aug represents the augmented Lagrangian function; λ represents the slack variable of the channel boundary constraint; μ represents the penalty coefficient;

[0041] The parallelized feasible direction method is used to solve the optimization problem, and the control sequence and Lagrange multiplier are updated synchronously in each iteration.

[0042] The optimized control sequence is fed forward to the ship actuator, and the uncompensated model error is fed back to the next cycle optimization process.

[0043] Optionally, the specific process of synchronizing the motion state parameters of the actual ship and the virtual ship model in real time based on dynamic path planning, building an emergency decision tree in combination with an expert knowledge base, and verifying the feasibility of the emergency decision through Monte Carlo simulation is as follows:

[0044] Real-time synchronization of motion state parameters of the actual ship and the virtual ship model through time-sensitive network protocol;

[0045] An expert knowledge base covering three typical scenarios, namely, ship loss of control, sudden obstacle approach, and hydrological mutation, was constructed. An event-driven decision tree generation mechanism was adopted, with the deviation of the ship's motion state, the approach rate of the channel obstacle, and the abnormal amplitude of the hydrological sensor as decision triggers. The risk level was quantified using a fuzzy membership function.

[0046] A three-layer decision tree structure is established based on expert experience rule sets. The first-layer nodes activate corresponding emergency modes based on risk types. The second-layer nodes calculate the feasible avoidance domain based on the ship's power redundancy and the maneuverable space in the channel. The third-layer nodes generate a complex maneuvering instruction set including emergency braking, Z-shaped maneuvers, and coordinated yielding.

[0047] A set of random interference variables, including sudden changes in water turbulence intensity, fluctuations in communication delays, and deviations in other ship trajectories, is injected into the virtual simulation platform. A Monte Carlo simulation is performed to simulate tens of thousands of scenarios. The success rate, path deviation, and energy consumption index of each emergency strategy are recorded, and the strategy set that meets the reliability threshold of the following formula (6) is selected:

[0048]

[0049] Among them, P success represents the success probability of the strategy; N represents the total number of simulations; I represents the indicative function; S i represents the safety constraint satisfaction event in the simulation; ||ΔX|| represents the final position deviation norm; ∈ represents the maximum allowable deviation threshold; P threshold Indicates the preset probability threshold;

[0050] The verified emergency strategies are encoded into executable instruction chains, and a strategy priority mapping table is established. When the matching degree between the actual navigation environment parameters and the simulation conditions exceeds the preset similarity threshold, the emergency decision-making plan of the corresponding level is automatically activated.

[0051] Optionally, the specific process of using edge computing nodes to perform local route optimization on dynamic path planning, performing multi-vessel trajectory prediction through a federated learning mechanism, and generating corresponding collaborative collision avoidance strategies is as follows:

[0052] Deploy edge computing nodes at key points along the waterway and on ships. Each node receives real-time sensor data from the ship itself and status information broadcast by neighboring ships, building a local situational awareness network.

[0053] Ship trajectory segments are extracted based on a spatiotemporal sliding window mechanism, a ship trajectory prediction model is constructed using a gated recurrent unit network, and a federated learning framework is used for distributed model training.

[0054] Each edge node uses local historical trajectory data to train the ship trajectory prediction model and generate model parameter increments;

[0055] The parameter increments are homomorphically encrypted through a trusted execution environment and uploaded to the cloud parameter aggregation server;

[0056] The dynamic weighted average algorithm is used to aggregate the global model parameters, as shown in the following formula (7):

[0057]

[0058] Among them, ω global represents the global model parameters after aggregation; M represents the number of participating nodes; α i Represents the dynamic weight of the i-th node; N i represents the size of the local dataset; Δt i and Δt j represents the data timeliness deviation; β represents the time decay factor;

[0059] The updated global model is distributed to each edge node, and multi-step trajectory prediction is performed synchronously to generate a probabilistic trajectory distribution map with confidence intervals.

[0060] A collaborative collision avoidance decision model based on potential field game is established, and the predicted trajectory of each ship is mapped into a time-varying potential field function, as shown in the following formula (8):

[0061]

[0062] Among them, U coll represents the collision potential field strength; K represents the potential field gain coefficient; d ij (t) represents the real-time distance between ship i and ship j; ξ represents the field intensity gradient factor;

[0063] Solve the Nash equilibrium point locally at the edge node to generate a set of cooperative heading adjustment values that satisfies the following formula (9):

[0064]

[0065] Where Δψ i represents the course correction of ship i; ψ i represents the current heading angle of the ship; η represents the potential field weight coefficient; X i Indicates the current position of the ship; v i represents the ship speed; Δt represents the decision cycle.

[0066] Optionally, a dynamic priority scheduling mechanism is established based on the collaborative collision avoidance strategy. When a navigation conflict or sudden hydrological change is detected, a graded response is implemented based on the ship's cargo level, draft, and power characteristics. The specific process of confirming the global avoidance plan through a distributed consensus algorithm is as follows:

[0067] A dynamic priority assessment model is constructed that includes the weight coefficient of the ship's cargo level, the draft safety margin, and the dynamic characteristic response threshold. The priority base value is initialized based on the ship's certificate data and the real-time monitored draft, main engine power, and propeller redundancy.

[0068] When the sensor detects a navigation conflict warning signal or abnormal fluctuations in the hydrological sensor output exceed a preset threshold, the priority recalculation module is triggered. The priority base value of each ship is dynamically revised using the analytic hierarchy process, combining the conflict urgency index and the hydrological mutation influencing factor to generate a real-time priority sequence. The conflict urgency index is composed of relative speed, closest approach distance, and braking distance margin; the hydrological mutation influencing factor includes flow velocity increment, flow direction deflection angle, and water depth sudden change amplitude.

[0069] Establish a hierarchical response strategy library, defining that high-priority ships have priority in keeping their routes, while low-priority ships have the obligation to actively avoid them. Different levels correspond to different course adjustment limits, speed gradient constraints, and avoidance execution time windows.

[0070] Each ship broadcasts an information package containing real-time priority, current motion status and feasible avoidance plan sets through the self-organizing network, and the shore-based control center simultaneously injects channel boundary constraints and real-time hydrological data;

[0071] An improved Byzantine fault-tolerant consensus algorithm is used to verify the compatibility of each ship's avoidance plan in a distributed node network. The optimal avoidance plan set that satisfies global navigation safety constraints is solved through iterative optimization. When more than two-thirds of the nodes reach a consensus, the global avoidance plan is confirmed and broadcast to relevant ships for execution.

[0072] Establish a priority time-attenuation mechanism to reduce the priority adjustment range in real time according to the conflict resolution time and hydrological stability status. When the system returns to normal navigation conditions, the priority parameters are reset to the initial state.

[0073] Based on the same inventive concept, the present invention also provides an intelligent navigation and emergency decision-making system for ships in complex waterways, which is used to implement the intelligent navigation and emergency decision-making method for ships in complex waterways, including a data acquisition and processing module, an environment modeling module, a path planning module, an emergency decision-making module, an edge computing and collaboration module, and a priority scheduling and consensus module connected in sequence;

[0074] The data acquisition and processing module is used to collect environmental data and perform multi-sensor spatiotemporal calibration and data compensation through an adaptive Kalman filter algorithm;

[0075] The environmental modeling module is used to perform grid division and voxel processing on the calibrated and compensated data, and generate and update the three-dimensional dynamic environmental model of the waterway in real time;

[0076] The path planning module is used to establish a multi-objective optimization function including ship maneuverability constraints, channel boundary constraints and navigation rules based on the ship kinematic model and real-time hydrodynamic parameters, and adopts an improved model predictive control algorithm for dynamic path planning;

[0077] The emergency decision module is used to synchronize the motion state parameters of the physical ship and the virtual model in real time, build an emergency decision tree based on the expert knowledge base, and verify the feasibility of the emergency decision through Monte Carlo simulation;

[0078] The edge computing and collaboration module is used to deploy edge computing nodes for local route optimization, perform multi-vessel trajectory prediction through a federated learning mechanism, and generate corresponding collaborative collision avoidance strategies;

[0079] The priority scheduling and consensus module is used to establish a dynamic priority scheduling mechanism. When navigation conflicts or sudden hydrological changes are detected, a graded response is implemented based on the ship's cargo level, draft depth and power characteristics, and a global avoidance plan is confirmed through a distributed consensus algorithm.

[0080] Based on the same inventive concept, the present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned complex waterway ship intelligent navigation and emergency decision-making method.

[0081] Based on the same inventive concept, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned intelligent navigation and emergency decision-making method for ships in complex waterways is implemented.

[0082] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0083] By collecting environmental data from a multi-source heterogeneous sensor array and combining it with an adaptive Kalman filter algorithm for multi-sensor spatiotemporal calibration and data compensation, this approach overcomes the existing system's reliance on a single sensor, which is susceptible to interference and suffers from low multi-source data fusion accuracy. This method fully integrates the advantages of different sensor types, effectively filtering out noise and interference, enabling the precise collection and processing of waterway environmental data, thereby establishing a complete and accurate three-dimensional dynamic environmental model of the waterway. This provides comprehensive, real-time, and reliable environmental information support for intelligent navigation and emergency decision-making, enabling ships to more clearly perceive the complex surrounding navigation environment, including narrow and winding waterways, densely packed ships, and complex hydrological conditions, laying a solid foundation for subsequent decision-making and planning.

[0084] Based on the ship's kinematic model and real-time hydrodynamic parameters, a multi-objective optimization function is established that includes ship maneuverability constraints, channel boundary constraints, and navigation rules, and an improved model predictive control algorithm is used for dynamic path planning. Compared with traditional path planning algorithms that are difficult to deal with the problem of dynamic avoidance and hydrological coupling constraints, the present invention can fully consider the physical characteristics of the ship itself, the actual limitations of the channel, and the requirements of navigation rules, and optimize the ship's navigation path in real time in a dynamically changing navigation environment. Not only can it achieve safe avoidance of other ships, but it can also combine real-time hydrological data to reasonably adjust the navigation route and avoid navigation risks caused by hydrological factors such as water currents and tides. This dynamic path planning method improves the navigation efficiency of ships in complex waterways, while significantly enhancing navigation safety and reducing the probability of accidents such as collisions and groundings.

[0085] The system synchronizes the motion parameters of the actual ship and the virtual ship model in real time, builds an emergency decision tree based on an expert knowledge base, and verifies the feasibility of emergency decisions through Monte Carlo simulation. This process overcomes the existing system's reliance on human experience for emergency response, transforming the expert's extensive experience into a quantifiable and verifiable decision model. When a ship encounters an emergency, the emergency decision tree quickly generates multiple possible emergency decision plans based on the ship's real-time motion status and environmental information. These plans are then simulated and verified through Monte Carlo simulation to select the most feasible and safest emergency measures. This scientific emergency decision-making mechanism enables accurate responses in the shortest possible time, improves the ship's autonomous decision-making capabilities in emergency situations, minimizes accident losses, and ensures the safety of the ship and personnel.

[0086] Edge computing nodes are used for local route optimization. A federated learning mechanism is used to predict multi-vessel trajectories and generate corresponding coordinated collision avoidance strategies. A dynamic priority scheduling mechanism is also established, implementing a graded response based on vessel cargo level, draft, and propulsion characteristics. A distributed consensus algorithm is used to confirm the global avoidance plan. These technical measures effectively address the existing system's issues of high ship-to-shore communication latency and insufficient global coordinated collision avoidance and resource scheduling capabilities. Edge computing nodes enable real-time processing and rapid decision-making of local vessel navigation data, improving the efficiency of local route optimization. The federated learning mechanism enables the sharing of trajectory prediction information among multiple vessels, enabling more accurate coordinated collision avoidance. Dynamic priority scheduling and a distributed consensus algorithm ensure that, in the event of navigation conflicts or sudden hydrological changes, graded responses are implemented based on the actual conditions of the vessels, allowing for the rapid formation of a globally consistent avoidance plan. This collaborative mechanism enables efficient communication and collaboration between ships and between ships and shore, optimizing the utilization of waterway resources, improving overall waterway efficiency and safety, and making vessel traffic management in complex waterways more organized and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 Schematic diagram of the process of intelligent navigation and emergency decision-making method for ships in complex waterways according to an embodiment of the present invention;

[0088] Figure 2 This is a structural diagram of the intelligent navigation and emergency decision-making system for ships in complex waterways according to an embodiment of the present invention. DETAILED DESCRIPTION

[0089] The following is a specific implementation method with reference to the accompanying drawings.

[0090] Reference Figure 1 , a method for intelligent navigation and emergency decision-making of ships in complex waterways, comprising the following steps:

[0091] Step 1: Collect environmental data through a multi-source heterogeneous sensor array, use the adaptive Kalman filter algorithm to perform multi-sensor spatiotemporal calibration and data compensation, and establish a three-dimensional dynamic environment model of the waterway.

[0092] In some embodiments, the specific process of collecting environmental data through a multi-source heterogeneous sensor array and using an adaptive Kalman filter algorithm to perform multi-sensor spatiotemporal calibration and data compensation to establish a three-dimensional dynamic environment model of the waterway is as follows:

[0093] A multi-source heterogeneous sensor array consisting of radar, lidar, visual camera, water depth sensor and ship automatic identification system is arranged at the bow, stern and both sides of the ship to collect channel boundaries, obstacle locations, ship motion parameters, water depth information and the identity status data of navigable ships;

[0094] Perform time stamp synchronization and spatial coordinate system conversion on the output data of each sensor to establish a unified time and space benchmark;

[0095] An adaptive adjustment model is constructed that includes the sensor measurement noise covariance matrix and the system process noise covariance matrix. The matrix parameters are dynamically adjusted according to the real-time environmental interference intensity. The measurement data of different sensors are iteratively processed using Kalman filtering to achieve spatiotemporal calibration and abnormal data compensation.

[0096] The calibrated and compensated static channel boundary data, dynamic ship motion data and real-time hydrological data are gridded and voxelized to generate a three-dimensional dynamic environment model of the channel that includes channel water depth, three-dimensional coordinates of obstacles and dynamic trajectory of ships. The model is refreshed in real time through the data update interface.

[0097] In some embodiments, the specific process of constructing the adaptive adjustment model including the sensor measurement noise covariance matrix and the system process noise covariance matrix is as follows:

[0098] Initialize the basic values of the sensor measurement noise covariance matrix and the system process noise covariance matrix based on the sensor's historical measurement data and the typical environmental parameters of the waterway;

[0099] The environmental interference monitoring module deployed in the sensor array collects multi-dimensional interference parameters including electromagnetic interference intensity, water turbidity and meteorological noise in real time, and constructs an interference intensity evaluation function.

[0100] A dynamic adjustment rule base for the noise covariance matrix is established. Based on the real-time interference intensity assessment results, the noise variance terms of each sensor's corresponding measurement dimension in the sensor measurement noise covariance matrix and the process noise terms of the ship kinematic model state transition in the system process noise covariance matrix are corrected through a fuzzy logic controller or an adaptive weight allocation algorithm.

[0101] The adjusted sensor measurement noise covariance matrix and system process noise covariance matrix are input into the Kalman filter iterative process to form a closed-loop feedback regulation mechanism to perform real-time adaptation to different sensor measurement noises and system dynamic noises.

[0102] Step 2: Based on the ship kinematic model and real-time hydrodynamic parameters, combined with the three-dimensional dynamic environment model of the channel, a multi-objective optimization function including ship maneuverability constraints, channel boundary constraints and navigation rules is established, and an improved model predictive control algorithm is used for dynamic path planning.

[0103] In some embodiments, based on the ship kinematic model and real-time hydrodynamic parameters, combined with the three-dimensional dynamic environment model of the channel, a multi-objective optimization function including ship maneuverability constraints, channel boundary constraints, and navigation rules is established. The specific process of dynamic path planning using the improved model predictive control algorithm is as follows:

[0104] The three-degree-of-freedom kinematic differential equation of the ship is established as shown in the following equation (1):

[0105]

[0106] Where (x, y) represents the position coordinates of the ship in the plane coordinate system; represents the velocity component of the ship along the x-axis; represents the velocity component of the ship along the y-axis; represents the ship's turning rate; ψ represents the ship's heading angle; u represents the ship's longitudinal speed along the stern direction of the hull; v represents the ship's transverse speed perpendicular to the stern direction of the hull; r represents the rate of change of the ship's heading angle; θ h represents the real-time hydrodynamic parameter set; Δ x (θ h ) represents the additional velocity component caused by the real-time hydrodynamic parameters in the x direction of the inertial coordinate system; Δ y (θ h ) represents the additional velocity component caused by the real-time hydrodynamic parameters in the y direction of the inertial coordinate system; Δ ψ (θ h ) represents the correction term for the rate of change of the ship's heading angle, which is used to compensate for the interference of hydrodynamic parameters on the ship's steering characteristics;

[0107] Construct a multi-objective optimization function as shown in the following formula (2):

[0108]

[0109] Among them, J represents the total cost function; N p Indicates the length of the prediction time domain; X k represents the ship state vector at step k; X ref,k represents the reference trajectory state vector of the kth step; Q represents the state tracking error weight matrix; U k represents the control input vector of the step; R represents the control input weight matrix; γ represents the obstacle repulsion coefficient; Ω k represents the state vector of the channel obstacle at step k; d(X k ,Ω k ) represents the minimum Euclidean distance between the k-th step ship position and the channel obstacle; ρ represents the speed limit penalty factor; V lim Indicates the speed limit value of the channel; v krepresents the actual speed of the ship at step k;

[0110] Define the time-varying constraint set as shown in the following formula (3):

[0111]

[0112] Where U(t) represents the control input variable of the ship at time t; U max (t) and U min (t) represents the maximum value vector and minimum value vector of the control input respectively; (x k ,y k ) represents the plane coordinate of the ship at the kth moment in the prediction time domain; represents the channel boundary constraint function; a gradient representing the channel geometry; Represents the position coordinate vector of the i-th obstacle; R safe represents the static safety radius; τ resp Indicates the braking response time;

[0113] The improved model predictive control algorithm is used to solve the constrained optimization function in the receding horizon, as shown in the following formula (4):

[0114]

[0115] Among them, f(X k ,U k ,θ h ) represents the state transfer function of the ship kinematic model, which is obtained by discretizing the differential equation of formula (1); X k+1 represents the ship state vector at step k+1;

[0116] Update θ through real-time hydrodynamic parameter identification module h The estimated value of is used to modify the ship kinematic model;

[0117] The augmented Lagrangian method is used to deal with the channel boundary constraints, as shown in the following formula (5):

[0118]

[0119] Among them, J aug represents the augmented Lagrangian function; λ represents the slack variable of the channel boundary constraint; μ represents the penalty coefficient;

[0120] The parallelized feasible direction method is used to solve the optimization problem, and the control sequence and Lagrange multiplier are updated synchronously in each iteration.

[0121] The optimized control sequence is fed forward to the ship actuator, and the uncompensated model error is fed back to the next cycle optimization process.

[0122] Step 3: Based on dynamic path planning, the motion state parameters of the actual ship and the virtual ship model are synchronized in real time, and an emergency decision tree is constructed in combination with the expert knowledge base. The feasibility of the emergency decision is verified through Monte Carlo simulation.

[0123] In some embodiments, based on dynamic path planning, the motion state parameters of the actual ship and the virtual ship model are synchronized in real time, and an emergency decision tree is constructed in combination with an expert knowledge base. The specific process of verifying the feasibility of the emergency decision through Monte Carlo simulation is as follows:

[0124] Real-time synchronization of motion state parameters of the actual ship and the virtual ship model through time-sensitive network protocol;

[0125] An expert knowledge base covering three typical scenarios, namely, ship loss of control, sudden obstacle approach, and hydrological mutation, was constructed. An event-driven decision tree generation mechanism was adopted, with the deviation of the ship's motion state, the approach rate of the channel obstacle, and the abnormal amplitude of the hydrological sensor as decision triggers. The risk level was quantified using a fuzzy membership function.

[0126] A three-layer decision tree structure is established based on expert experience rule sets. The first-layer nodes activate corresponding emergency modes based on risk types. The second-layer nodes calculate the feasible avoidance domain based on the ship's power redundancy and the maneuverable space in the channel. The third-layer nodes generate a complex maneuvering instruction set including emergency braking, Z-shaped maneuvers, and coordinated yielding.

[0127] A set of random interference variables, including sudden changes in water turbulence intensity, fluctuations in communication delays, and deviations in other ship trajectories, is injected into the virtual simulation platform. A Monte Carlo simulation is performed to simulate tens of thousands of scenarios. The success rate, path deviation, and energy consumption index of each emergency strategy are recorded, and the strategy set that meets the reliability threshold of the following formula (6) is selected:

[0128]

[0129] Among them, P success represents the success probability of the strategy; N represents the total number of simulations; I represents the indicative function; S i represents the safety constraint satisfaction event in the simulation; ||ΔX|| represents the final position deviation norm; ∈ represents the maximum allowable deviation threshold; P threshold Indicates the preset probability threshold;

[0130] The verified emergency strategies are encoded into executable instruction chains, and a strategy priority mapping table is established. When the matching degree between the actual navigation environment parameters and the simulation conditions exceeds the preset similarity threshold, the emergency decision-making plan of the corresponding level is automatically activated.

[0131] Step 4: Use edge computing nodes to perform local route optimization for dynamic path planning, predict multi-ship trajectories through the federated learning mechanism, and generate corresponding collaborative collision avoidance strategies.

[0132] In some embodiments, the specific process of using edge computing nodes to perform local route optimization for dynamic path planning, predicting multiple ship trajectories through a federated learning mechanism, and generating corresponding collaborative collision avoidance strategies is as follows:

[0133] Deploy edge computing nodes at key points along the waterway and on ships. Each node receives real-time sensor data from the ship itself and status information broadcast by neighboring ships, building a local situational awareness network.

[0134] Ship trajectory segments are extracted based on a spatiotemporal sliding window mechanism, a ship trajectory prediction model is constructed using a gated recurrent unit network, and a federated learning framework is used for distributed model training.

[0135] Each edge node uses local historical trajectory data to train the ship trajectory prediction model and generate model parameter increments;

[0136] The parameter increments are homomorphically encrypted through a trusted execution environment and uploaded to the cloud parameter aggregation server;

[0137] The dynamic weighted average algorithm is used to aggregate the global model parameters, as shown in the following formula (7):

[0138]

[0139] Among them, ω global represents the global model parameters after aggregation; M represents the number of participating nodes; α i Represents the dynamic weight of the i-th node; N i represents the size of the local dataset; Δt i and Δt j represents the data timeliness deviation; β represents the time decay factor;

[0140] The updated global model is distributed to each edge node, and multi-step trajectory prediction is performed synchronously to generate a probabilistic trajectory distribution map with confidence intervals.

[0141] A collaborative collision avoidance decision model based on potential field game is established, and the predicted trajectory of each ship is mapped into a time-varying potential field function, as shown in the following formula (8):

[0142]

[0143] Among them, U coll represents the collision potential field strength; K represents the potential field gain coefficient; d ij (t) represents the real-time distance between ship i and ship j; ξ represents the field intensity gradient factor;

[0144] Solve the Nash equilibrium point locally at the edge node to generate a set of cooperative heading adjustment values that satisfies the following formula (9):

[0145]

[0146] Where Δψ i represents the course correction of ship i; ψ i represents the current heading angle of the ship; η represents the potential field weight coefficient; X i Indicates the current position of the ship; v i represents the ship speed; Δt represents the decision cycle.

[0147] Step 5. Based on the collaborative collision avoidance strategy, a dynamic priority scheduling mechanism is established. When navigation conflicts or sudden hydrological changes are detected, a graded response is implemented based on the ship's cargo level, draft depth, and power characteristics, and a global avoidance plan is confirmed through a distributed consensus algorithm.

[0148] In some embodiments, a dynamic priority scheduling mechanism is established based on the collaborative collision avoidance strategy. When a navigation conflict or sudden hydrological change is detected, a graded response is implemented based on the ship's cargo level, draft, and power characteristics. The specific process of confirming the global avoidance plan through a distributed consensus algorithm is as follows:

[0149] A dynamic priority assessment model is constructed that includes the weight coefficient of the ship's cargo level, the draft safety margin, and the dynamic characteristic response threshold. The priority base value is initialized based on the ship's certificate data and the real-time monitored draft, main engine power, and propeller redundancy.

[0150] When the sensor detects a navigation conflict warning signal or abnormal fluctuations in the hydrological sensor output exceed a preset threshold, the priority recalculation module is triggered. Combining the conflict urgency index with the hydrological mutation influencing factor, the hierarchical analysis method dynamically modifies the priority base value of each ship to generate a real-time priority sequence. The conflict urgency index is composed of relative speed, closest approach distance, and braking distance margin; the hydrological mutation influencing factor includes flow velocity increment, flow direction deflection angle, and water depth sudden change amplitude.

[0151] Establish a hierarchical response strategy library, defining that high-priority ships have priority in keeping their routes, while low-priority ships have the obligation to actively avoid them. Different levels correspond to different course adjustment limits, speed gradient constraints, and avoidance execution time windows.

[0152] Each ship broadcasts an information package containing real-time priority, current motion status and feasible avoidance plan sets through the self-organizing network, and the shore-based control center simultaneously injects channel boundary constraints and real-time hydrological data;

[0153] An improved Byzantine fault-tolerant consensus algorithm is used to verify the compatibility of each ship's avoidance plan in a distributed node network. The optimal avoidance plan set that satisfies global navigation safety constraints is solved through iterative optimization. When more than two-thirds of the nodes reach a consensus, the global avoidance plan is confirmed and broadcast to relevant ships for execution.

[0154] A priority time-attenuation mechanism is established to reduce the priority adjustment range in real time according to the conflict resolution time and hydrological stability status. When the system returns to normal navigation conditions, the priority parameters are reset to the initial state.

[0155] Based on the same inventive concept, corresponding to any of the above embodiments, refer to Figure 2 The present invention provides an intelligent navigation and emergency decision-making system for ships in complex waterways, which is used to implement the aforementioned intelligent navigation and emergency decision-making method for ships in complex waterways, including a data acquisition and processing module, an environment modeling module, a path planning module, an emergency decision-making module, an edge computing and collaboration module, and a priority scheduling and consensus module connected in sequence;

[0156] The data acquisition and processing module is used to collect environmental data and perform multi-sensor spatiotemporal calibration and data compensation through an adaptive Kalman filter algorithm;

[0157] The environmental modeling module is used to perform grid division and voxel processing on the calibrated and compensated data, and generate and update the three-dimensional dynamic environmental model of the waterway in real time;

[0158] The path planning module is used to establish a multi-objective optimization function including ship maneuverability constraints, channel boundary constraints and navigation rules based on the ship kinematic model and real-time hydrodynamic parameters, and adopts an improved model predictive control algorithm for dynamic path planning;

[0159] The emergency decision module is used to synchronize the motion state parameters of the physical ship and the virtual model in real time, build an emergency decision tree based on the expert knowledge base, and verify the feasibility of the emergency decision through Monte Carlo simulation;

[0160] The edge computing and collaboration module is used to deploy edge computing nodes for local route optimization, perform multi-vessel trajectory prediction through a federated learning mechanism, and generate corresponding collaborative collision avoidance strategies;

[0161] The priority scheduling and consensus module is used to establish a dynamic priority scheduling mechanism. When navigation conflicts or sudden hydrological changes are detected, a graded response is implemented based on the ship's cargo level, draft depth and power characteristics, and a global avoidance plan is confirmed through a distributed consensus algorithm.

[0162] Based on the same inventive concept, corresponding to any of the above embodiments, the present invention provides an electronic device, including a memory and a processor, the memory is used to store computer programs, and the processor runs the computer programs to enable the electronic device to execute the complex waterway ship intelligent navigation and emergency decision-making method of the embodiment.

[0163] Optionally, the above-mentioned electronic device may be a server.

[0164] In addition, this embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the complex waterway ship intelligent navigation and emergency decision-making method of the embodiment is implemented.

[0165] It is understood that the processor in the embodiments of the present invention may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0166] The method steps in the embodiments of the present invention can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an ASIC.

[0167] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a storage medium or transmitted via a storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

Claims

1. A method for intelligent navigation and emergency decision-making of ships in complex waterways, characterized by: The following steps are involved: Environmental data is collected through a multi-source heterogeneous sensor array, and an adaptive Kalman filter algorithm is used to perform multi-sensor spatiotemporal calibration and data compensation to establish a three-dimensional dynamic environment model of the waterway. Based on the ship's kinematic model and real-time hydrodynamic parameters, combined with the three-dimensional dynamic environment model of the channel, a multi-objective optimization function is established that includes ship maneuverability constraints, channel boundary constraints, and navigation rules. An improved model predictive control algorithm is used for dynamic path planning. Based on dynamic path planning, the motion state parameters of the actual ship and the virtual ship model are synchronized in real time. An emergency decision tree is constructed in combination with the expert knowledge base, and the feasibility of emergency decision-making is verified through Monte Carlo simulation. Edge computing nodes are used to perform local route optimization for dynamic path planning, and a federated learning mechanism is used to predict multi-ship trajectories and generate corresponding collaborative collision avoidance strategies. Based on the collaborative collision avoidance strategy, a dynamic priority scheduling mechanism is established. When navigation conflicts or sudden hydrological changes are detected, a graded response is implemented based on the ship's cargo level, draft and power characteristics, and the global avoidance plan is confirmed through a distributed consensus algorithm.

2. The method for intelligent navigation and emergency decision-making of ships in complex waterways according to claim 1, characterized in that: The specific process of collecting environmental data through a multi-source heterogeneous sensor array, using an adaptive Kalman filter algorithm to perform multi-sensor spatiotemporal calibration and data compensation, and establishing a three-dimensional dynamic environment model of the waterway is as follows: A multi-source heterogeneous sensor array consisting of radar, lidar, visual camera, water depth sensor and ship automatic identification system is arranged at the bow, stern and both sides of the ship to collect channel boundaries, obstacle locations, ship motion parameters, water depth information and the identity status data of navigable ships; Perform time stamp synchronization and spatial coordinate system conversion on the output data of each sensor to establish a unified time and space benchmark; An adaptive adjustment model is constructed that includes the sensor measurement noise covariance matrix and the system process noise covariance matrix. The matrix parameters are dynamically adjusted according to the real-time environmental interference intensity. The measurement data of different sensors are iteratively processed using Kalman filtering to achieve spatiotemporal calibration and abnormal data compensation. The calibrated and compensated static channel boundary data, dynamic ship motion data and real-time hydrological data are gridded and voxelized to generate a three-dimensional dynamic environment model of the channel that includes channel water depth, three-dimensional coordinates of obstacles and dynamic trajectory of ships. The model is refreshed in real time through the data update interface.

3. The method for intelligent navigation and emergency decision-making of ships in complex waterways according to claim 2, characterized in that: The specific process of constructing the adaptive adjustment model including the sensor measurement noise covariance matrix and the system process noise covariance matrix is as follows: Initialize the basic values of the sensor measurement noise covariance matrix and the system process noise covariance matrix based on the sensor's historical measurement data and the typical environmental parameters of the waterway; The environmental interference monitoring module deployed in the sensor array collects multi-dimensional interference parameters including electromagnetic interference intensity, water turbidity and meteorological noise in real time, and constructs an interference intensity evaluation function. A dynamic adjustment rule base for the noise covariance matrix is established. Based on the real-time interference intensity assessment results, the noise variance terms of each sensor's corresponding measurement dimension in the sensor measurement noise covariance matrix and the process noise terms of the ship kinematic model state transition in the system process noise covariance matrix are corrected through a fuzzy logic controller or an adaptive weight allocation algorithm. The adjusted sensor measurement noise covariance matrix and system process noise covariance matrix are input into the Kalman filter iterative process to form a closed-loop feedback regulation mechanism to adapt different sensor measurement noises and system dynamic noise in real time.

4. The method for intelligent navigation and emergency decision-making of ships in complex waterways according to claim 1, characterized in that: The specific process of establishing a multi-objective optimization function including ship maneuverability constraints, channel boundary constraints and navigation rules based on the ship kinematic model and real-time hydrodynamic parameters in combination with the three-dimensional dynamic environment model of the channel and using the improved model predictive control algorithm for dynamic path planning is as follows: The three-degree-of-freedom kinematic differential equation of the ship is established as shown in the following equation (1): Where (x, y) represents the position coordinates of the ship in the plane coordinate system; represents the velocity component of the ship along the x-axis; represents the velocity component of the ship along the y-axis; represents the ship's turning rate; ψ represents the ship's heading angle; u represents the ship's longitudinal speed along the stern direction of the hull; v represents the ship's transverse speed perpendicular to the stern direction of the hull; r represents the rate of change of the ship's heading angle; θ h represents the real-time hydrodynamic parameter set; Δ x (θ h ) represents the additional velocity component caused by the real-time hydrodynamic parameters in the x direction of the inertial coordinate system; Δ y (θ h ) represents the additional velocity component caused by the real-time hydrodynamic parameters in the y direction of the inertial coordinate system; Δ ψ (θ h ) represents the correction term for the rate of change of the ship's heading angle, which is used to compensate for the interference of hydrodynamic parameters on the ship's steering characteristics; Construct a multi-objective optimization function as shown in the following formula (2): Among them, J represents the total cost function; N p Indicates the length of the prediction time domain; X k represents the ship state vector at step k; X ref,k represents the reference trajectory state vector of the kth step; Q represents the state tracking error weight matrix; U k represents the control input vector of the step; R represents the control input weight matrix; γ represents the obstacle repulsion coefficient; Ω k represents the state vector of the channel obstacle at step k; d(X k ,Ω k ) represents the minimum Euclidean distance between the k-th step ship position and the channel obstacle; ρ represents the speed limit penalty factor; V lim Indicates the speed limit value of the channel; v k represents the actual speed of the ship at step k; Define the time-varying constraint set as shown in the following formula (3): Where U(t) represents the control input variable of the ship at time t; U max (t) and U min (t) represents the maximum value vector and minimum value vector of the control input respectively; (x k ,y k ) represents the plane coordinate of the ship at the kth moment in the prediction time domain; represents the channel boundary constraint function; a gradient representing the geometry of the channel; Represents the position coordinate vector of the i-th obstacle; R safe represents the static safety radius; τ resp Indicates the braking response time; The improved model predictive control algorithm is used to solve the constrained optimization function in the receding horizon, as shown in the following formula (4): Among them, f(X k ,U k ,θ h ) represents the state transfer function of the ship kinematic model, which is obtained by discretizing the differential equation of formula (1); X k+1 represents the ship state vector at step k+1; Update θ through real-time hydrodynamic parameter identification module h The estimated value of is used to modify the ship kinematic model; The augmented Lagrangian method is used to deal with the channel boundary constraints, as shown in the following formula (5): Among them, J aug represents the augmented Lagrangian function; λ represents the slack variable of the channel boundary constraint; μ represents the penalty coefficient; The parallelized feasible direction method is used to solve the optimization problem, and the control sequence and Lagrange multiplier are updated synchronously in each iteration. The optimized control sequence is fed forward to the ship actuator, and the uncompensated model error is fed back to the next cycle optimization process.

5. The method for intelligent navigation and emergency decision-making of ships in complex waterways according to claim 1, characterized in that: The specific process of synchronizing the motion state parameters of the actual ship and the virtual ship model in real time based on dynamic path planning, building an emergency decision tree in combination with the expert knowledge base, and verifying the feasibility of the emergency decision through Monte Carlo simulation is as follows: Real-time synchronization of motion state parameters of the actual ship and the virtual ship model through time-sensitive network protocol; An expert knowledge base covering three typical scenarios, namely, ship loss of control, sudden obstacle approach, and hydrological mutation, was constructed. An event-driven decision tree generation mechanism was adopted, with the deviation of the ship's motion state, the approach rate of the channel obstacle, and the abnormal amplitude of the hydrological sensor as decision triggers. The risk level was quantified using a fuzzy membership function. A three-layer decision tree structure is established based on expert experience rule sets. The first-layer nodes activate corresponding emergency modes based on risk types. The second-layer nodes calculate the feasible avoidance domain based on the ship's power redundancy and the maneuverable space in the channel. The third-layer nodes generate a complex maneuvering instruction set including emergency braking, Z-shaped maneuvers, and coordinated yielding. A set of random interference variables, including sudden changes in water turbulence intensity, fluctuations in communication delays, and deviations in other ship trajectories, is injected into the virtual simulation platform. A Monte Carlo simulation is performed to simulate tens of thousands of scenarios. The success rate, path deviation, and energy consumption index of each emergency strategy are recorded, and the strategy set that meets the reliability threshold of the following formula (6) is selected: Among them, P success represents the success probability of the strategy; N represents the total number of simulations; I represents the indicative function; S i represents the safety constraint satisfaction event in the simulation; ||ΔX|| represents the final position deviation norm; ∈ represents the maximum allowable deviation threshold; P threshold Indicates the preset probability threshold; The verified emergency strategies are encoded into executable instruction chains, and a strategy priority mapping table is established. When the matching degree between the actual navigation environment parameters and the simulation conditions exceeds the preset similarity threshold, the emergency decision-making plan of the corresponding level is automatically activated.

6. The method for intelligent navigation and emergency decision-making of ships in complex waterways according to claim 1, characterized in that: The specific process of using edge computing nodes to perform local route optimization for dynamic path planning, predicting multiple ship trajectories through a federated learning mechanism, and generating corresponding collaborative collision avoidance strategies is as follows: Edge computing nodes are deployed at key points along the waterway and on ships. Each node receives real-time sensor data from the ship itself and status information broadcast by neighboring ships, building a local situational awareness network. Ship trajectory segments are extracted based on a spatiotemporal sliding window mechanism, a ship trajectory prediction model is constructed using a gated recurrent unit network, and a federated learning framework is used for distributed model training. Each edge node uses local historical trajectory data to train the ship trajectory prediction model and generate model parameter increments; The parameter increments are homomorphically encrypted through a trusted execution environment and uploaded to the cloud parameter aggregation server; The dynamic weighted average algorithm is used to aggregate the global model parameters, as shown in the following formula (7): Among them, ω global represents the global model parameters after aggregation; M represents the number of participating nodes; α i Represents the dynamic weight of the i-th node; N i represents the size of the local dataset; Δt i and Δt j represents the data timeliness deviation; β represents the time decay factor; The updated global model is distributed to each edge node, and multi-step trajectory prediction is performed synchronously to generate a probabilistic trajectory distribution map with confidence intervals. A collaborative collision avoidance decision model based on potential field game is established, and the predicted trajectory of each ship is mapped into a time-varying potential field function, as shown in the following formula (8): Among them, U coll represents the collision potential field strength; K represents the potential field gain coefficient; d ij (t) represents the real-time distance between ship i and ship j; ξ represents the field intensity gradient factor; Solve the Nash equilibrium point locally at the edge node to generate a set of cooperative heading adjustment values that satisfies the following formula (9): Where Δψ i represents the course correction of ship i; ψ i represents the current heading angle of the ship; η represents the potential field weight coefficient; X i Indicates the current position of the ship; v i represents the ship speed; Δt represents the decision cycle.

7. The method for intelligent navigation and emergency decision-making of ships in complex waterways according to claim 1, characterized in that: According to the collaborative collision avoidance strategy, a dynamic priority scheduling mechanism is established. When a navigation conflict or sudden hydrological change is detected, a graded response is implemented based on the ship's cargo level, draft, and power characteristics. The specific process of confirming the global avoidance plan through a distributed consensus algorithm is as follows: A dynamic priority assessment model is constructed that includes the weight coefficient of the ship's cargo level, the draft safety margin, and the dynamic characteristic response threshold. The priority base value is initialized based on the ship's certificate data and the real-time monitored draft, main engine power, and propeller redundancy. When the sensor detects a navigation conflict warning signal or abnormal fluctuations in the hydrological sensor output exceed a preset threshold, the priority recalculation module is triggered. The priority base value of each ship is dynamically revised using the analytic hierarchy process, combining the conflict urgency index and the hydrological mutation influencing factor to generate a real-time priority sequence. The conflict urgency index is composed of relative speed, closest approach distance, and braking distance margin; the hydrological mutation influencing factor includes flow velocity increment, flow direction deflection angle, and water depth sudden change amplitude. Establish a hierarchical response strategy library, defining that high-priority ships have priority in keeping their routes, while low-priority ships have the obligation to actively avoid them. Different levels correspond to different course adjustment limits, speed gradient constraints, and avoidance execution time windows. Each ship broadcasts an information package containing real-time priority, current motion status and feasible avoidance plan sets through the self-organizing network, and the shore-based control center simultaneously injects channel boundary constraints and real-time hydrological data; An improved Byzantine fault-tolerant consensus algorithm is used to verify the compatibility of each ship's avoidance plan in a distributed node network. The optimal avoidance plan set that satisfies global navigation safety constraints is solved through iterative optimization. When more than two-thirds of the nodes reach a consensus, the global avoidance plan is confirmed and broadcast to relevant ships for execution. A priority time-attenuation mechanism is established to reduce the priority adjustment range in real time according to the conflict resolution time and hydrological stability status. When the system returns to normal navigation conditions, the priority parameters are reset to the initial state.

8. An intelligent navigation and emergency decision-making system for ships in complex waterways, used to implement the intelligent navigation and emergency decision-making method for ships in complex waterways according to any one of claims 1 to 7, characterized in that: It includes a data acquisition and processing module, an environment modeling module, a path planning module, an emergency decision-making module, an edge computing and collaboration module, and a priority scheduling and consensus module, which are connected in sequence; The data acquisition and processing module is used to collect environmental data and perform multi-sensor spatiotemporal calibration and data compensation through an adaptive Kalman filter algorithm; The environmental modeling module is used to perform grid division and voxel processing on the calibrated and compensated data, and generate and update the three-dimensional dynamic environmental model of the waterway in real time; The path planning module is used to establish a multi-objective optimization function including ship maneuverability constraints, channel boundary constraints and navigation rules based on the ship kinematic model and real-time hydrodynamic parameters, and adopts an improved model predictive control algorithm for dynamic path planning; The emergency decision module is used to synchronize the motion state parameters of the physical ship and the virtual model in real time, build an emergency decision tree based on the expert knowledge base, and verify the feasibility of the emergency decision through Monte Carlo simulation; The edge computing and collaboration module is used to deploy edge computing nodes for local route optimization, perform multi-vessel trajectory prediction through a federated learning mechanism, and generate corresponding collaborative collision avoidance strategies; The priority scheduling and consensus module is used to establish a dynamic priority scheduling mechanism. When navigation conflicts or sudden hydrological changes are detected, a graded response is implemented based on the ship's cargo level, draft depth and power characteristics, and a global avoidance plan is confirmed through a distributed consensus algorithm.

9. An electronic device, characterized in that: It includes a memory and a processor, the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the complex waterway ship intelligent navigation and emergency decision-making method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the complex waterway ship intelligent navigation and emergency decision-making method according to any one of claims 1 to 7.

Images