Research On Multi-vessel Collision Avoidance Based On Improved Adaptive Genetic Algorithm

Water transport is a very important mode of transportation.Against the backdrop of promoting energy conservation and emission reduction worldwide,water transport,as a relatively environmentally friendly mode of transportation,has received great attention.Further increasing the proportion of water transport in world transportation is particularly important to achieve the goal of carbon neutrality and peak carbon emissions.Currently,the greatest threat to water transport is ship collision accidents,and the main cause of such accidents is human error.Therefore,implementing intelligent navigation of ships can effectively improve the safety of water transport and increase its transportation efficiency.(1)The article proposes a multi-ship collision avoidance solution based on graded and strategic approaches,starting from two aspects: ship collision risk assessment and optimized collision avoidance strategies.By assessing the collision risk of the ships,the ships are divided into three navigation states: direct navigation,general collision avoidance,and emergency collision avoidance,and different collision avoidance strategies are adopted in each state.General collision avoidance uses the method of route planning based on improved adaptive genetic algorithm for multi-ship collision avoidance strategy;emergency collision avoidance uses the method of directly outputting heading and speed commands to quickly reduce collision risk,and the collision avoidance strategy is based on the velocity obstacle method.(2)This article proposes a method for evaluating the risk of ship collisions by using the length and width dimensions of the ship to establish a safety zone.The degree of intrusion and intrusion time of the speed obstacle cone formed by our ship and other ships are used as evaluation indicators,and the collision risk is quantified in terms of heading and speed,making the method linearly predictive and applicable to narrow water environments.The evaluation method was validated using real AIS historical data of ships in inland waterways,and compared with traditional evaluation methods.The experimental results show that the proposed evaluation method is more accurate and adaptable to restricted water areas than traditional evaluation methods.(3)This article proposes an improved adaptive genetic algorithm that optimizes chromosome encoding,population initialization methods,adaptive genetic operators,and elite preservation.The new chromosome encoding includes a speed change process and handles turning in accordance with ship kinematic constraints,ensuring the algorithm’s applicability in real ships.The fitness function includes a screening step that evaluates the overall genes of the population using survival rate as a parameter to control the number of offspring in the next generation.The Pareto sorting is used to achieve multi-objective sorting,and all Pareto first frontiers are preserved as elites to accelerate convergence.The concept of elite rate is introduced to represent the ratio of the number of Pareto first frontiers chromosomes to the total number of chromosomes in the population.The crossover operator uses survival rate to control the search pressure of the algorithm,while the mutation operator uses the elite rate to control the diversity of genes in the population and the number of iterations to control the convergence speed.This article also proposes a cumulative chromosome initialization method for the initial population to improve the convergence speed and optimization quality.This method improves the genetic structure of the initial population by accumulating good chromosomes,thus avoiding the high computational cost of searching during the adaptive initial stage.Finally,this article verifies the excellent performance of the initial accumulation,fast elite preservation,and adaptive genetic operators through comparative experiments.(4)This article uses the hardware information of the dolphin-16 unmanned boat as a reference for simulation experiments,and uses MATLAB to write simulation programs to verify the general collision avoidance and two-level collision avoidance strategies.The experimental results show that the general collision avoidance strategy can plan a more optimal solution that combines economy and safety in a limited time,and the two-level collision avoidance can further improve the reliability of general collision avoidance.