Research On Ship Path Planning Method Based On AIS Big Data

With the increase in maritime traffic,the density of maritime traffic flow continues to rise,making the maritime navigational environment and vessel route planning more complex.Route planning is an extremely important and tedious task for vessel voyage planning,aiming to provide vessels with the safest and most reasonable route to ensure the safety and cost-effectiveness of their voyages.The accuracy of route planning directly affects the safety of vessel navigation.Route planning is an integral part of voyage planning and has always been the primary issue to be addressed during vessel navigation.Traditional ship route planning methods include route planning based on paper charts and route planning based on electronic charts.However,the efficiency of paper chart route planning methods is low and reliability is poor,increasing the workload for ship navigators.In contrast,route planning based on electronic charts can automatically calculate route parameters such as distance and time,making it more efficient.However,fundamentally,route planning based on electronic charts does not differ significantly;it simply transfers the task of route planning from paper charts to electronic charts.With the development of ship intelligence,many intelligent ship route planning algorithms continuously emerge,which are more efficient and accurate compared to traditional route planning methods.However,most intelligent algorithms often suffer from the following issues: they often prioritize objectives such as the shortest route,minimum fuel consumption,and maximum economy,neglecting problems such as obstructive traffic flow direction and navigational habits that pose risks to navigation;over-reliance on electronic charts leads to unreliable planned routes due to electronic chart data errors or delayed updates;failure to consider factors closely related to ship routes such as ship type,tonnage,and size results in planned routes unsuitable for specific types of ships.Therefore,it is necessary to adopt more efficient and accurate route planning methods to ensure the safety and economy of ship navigation.AIS big data integrates various sensor information and collects a vast amount of vessel positions and voyage data.Each AIS data corresponds to specific vessel track information.By analyzing and studying vessel’s historical tracks,information such as vessel’s historical navigation practices and safe routes can be obtained,providing new insights for route planning research.By utilizing the comprehensive navigation environment information and traffic control management information provided by AIS,combined with intelligent processing techniques,route planning can reduce human errors and further enhance the safety,cost-effectiveness,and reliability of routes.This approach is significant in ensuring safe and economically efficient maritime navigation,improving shipping efficiency,reducing operational costs,minimizing maritime accidents,and mitigating the environmental impact of vessel transportation.Therefore,this paper proposes a vessel route planning method based on AIS big data,aiming to provide new perspectives for route planning research and improve the accuracy and reliability of route planning.It mainly consists of three parts: AIS big data trajectory feature extraction,AIS data clustering,and route optimization generation.1.AIS Trajectory Feature Extraction Based on Improved Sliding Window Algorithm(1)AIS Big Data Preprocessing: AIS big data preprocessing is an essential prerequisite for conducting route planning based on AIS big data,as it involves a significant amount of erroneous data that needs to be effectively eliminated and cleaned.To improve the accuracy and efficiency of data cleaning,this paper establishes an appropriate mathematical model of nautical charts and converts the latitude and longitude coordinates of trajectory points into position coordinates in the Mercator coordinate system.Considering the characteristics of AIS data,methods for removing duplicate data,detecting erroneous data,detecting missing data,and data completion based on polynomial interpolation are proposed.This paper collected over 5 million AIS trajectory data points from 64,586 vessel voyages in the northern waters of Zhoushan for 60 days and performed data cleaning.By comparing the macroscopic distribution situation and the dynamic movement of individual vessel trajectories,the feasibility and effectiveness of data cleaning were validated.The cleaned data eliminated erroneous data,corrected vessel position deviations,and removed some noise points.(2)Extraction of AIS Trajectory Feature Points.The AIS trajectory data is processed for key feature point extraction using an improved Sliding Window algorithm.Building upon the classical Sliding Window algorithm,this paper introduces additional parameters such as the turning angle threshold and modifies the window pointer movement method to achieve more precise trajectory feature extraction.By incorporating concepts from the domain of maritime navigation,an adaptive distance threshold for water areas is established,setting individual distance thresholds for each vessel trajectory.Additionally,the turning angle threshold is determined through mathematical statistics on AIS big data.Through algorithm comparison and experimental validation,it is demonstrated that the improved Sliding Window algorithm exhibits certain advantages.The feature points extracted by the improved Sliding Window algorithm retain the vessel’s turning points,effectively capturing the gradual turning process and enhancing computational efficiency.Moreover,they provide a better representation of the original vessel trajectory characteristics.2.AIS Data Clustering Based on Improved DBSCAN Algorithm(1)Improved DBSCAN Clustering Algorithm: The improved DBSCAN algorithm is proposed in this paper to better address the precise clustering of turning points in AIS trajectory data.Firstly,to address the problem of mixed clustering of different types of trajectory data,a ship classification method based on Random Forest is introduced to learn the attributes of the dataset,ensuring more accurate clustering of clusters.Secondly,by introducing the adaptive method of Firefly Algorithm,the problem of manually setting the parameters Eps and Min Pts of the DBSCAN algorithm is solved,thus improving the clustering effect.Ship AIS trajectory data from the northern waters of Zhoushan is tested,and a Random Forest classification algorithm is used to classify four types of ships: fishing boats,passenger ships,cargo ships,and oil tankers.The results show that the classification recognition accuracies of the four types of ships are 90.6%,94.2%,91.5%,and 88.7%,respectively.By comparing seven evaluation indicators including clustering accuracy(Accuracy),F-measure value(F),Rand index(RI),adjusted Rand index(ARI),mutual information-based score(MI),adjusted mutual information(AMI),and normalized mutual information(NMI)with traditional clustering algorithms,the superiority of the improved DBSCAN algorithm proposed in this paper is demonstrated.(2)Construction of Directed Graph for Navigational Routes: Since the turning points of AIS trajectories obtained from clustering may belong to different ships and leave trajectory points according to different headings,it is necessary to determine the connectivity of the turning points.This paper proposes 9 rules to determine the connectivity between two points.Finally,a directed network of shipping routes is constructed based on the turning points and their connectivity,serving as the basis for the next step of route optimization.3.Ship Path Planning Based on Slime Mold-Quantum Particle Swarm Optimization Fusion Algorithm(1)Slime Mold-Quantum Particle Swarm Optimization Fusion Algorithm: Slime mold algorithm,particle swarm algorithm,and their improved versions have been widely used in the field of optimization.The slime mold algorithm has strong optimization capabilities,but its systematicity is weak,and it is prone to local optima and post-iteration oscillation.On the other hand,the quantum particle swarm optimization algorithm,which improves the evolutionary search strategy of the particle swarm algorithm,has the advantages of fewer control parameters and faster computation speed.However,it tends to get trapped in local optima and exhibit premature convergence characteristics in densely populated areas of path points.To integrate the advantages of both algorithms,this paper proposes a new Slime Mold-Quantum Particle Swarm Optimization Fusion Algorithm(SMQPSOA)to improve the speed and quality of path optimization.The algorithm first uses the slime mold algorithm to select high-quality route segments and then uses the endpoints of these segments as fixed points to guide the directional selection of the quantum particle swarm optimization algorithm.This reduces the number of visited path points,decreases computational complexity,and improves search efficiency.(2)Simulation Results and Analysis: The effectiveness of the proposed SMQPSOA algorithm is verified through simulation experiments.According to the simulation results,the path length and average path length of the SMQPSOA algorithm are superior to the other two algorithms.Performance metrics such as path length standard deviation,average λ-branching factor,and convergence iteration count also show significant improvement.This method not only avoids the algorithm getting trapped in local optima and increases the probability of selecting the optimal path but also reduces the number of path points,decreases algorithm fluctuations,enhances robustness,and effectively improves search efficiency.This demonstrates the superiority of the proposed fusion algorithm over the original algorithms in terms of path optimization.(3)Analysis of Ship Path Planning Examples: To validate the superiority of the method proposed in this paper in ship route planning,different ship voyage tasks in obstructed water areas and open water areas were selected for ship route planning.A comparison study was conducted with routes planned by the A* algorithm and the improved Ant Colony Optimization(ACO)algorithm.The results show that the routes planned by the method in this paper are more effective in avoiding obstacles,meeting ship draft requirements,complying with local regulatory constraints,satisfying international collision avoidance rules,and adhering to ship traffic flow directions and navigational habits,thus performing better in terms of ship safety.The planned routes effectively avoid these potential navigational risks.It can be seen that the ship route planning method designed in this paper is not only reasonable and feasible but also has obvious superiority among similar ship route planning methods.