Research On Motion Planning Of Underwater Vehicles For Marine Organism Capture

With the development of underwater vehicle operation technology,more and more complex tasks are undertaken by underwater vehicles,and autonomous operation of underwater vehicle has become a research hotspot and future direction.Research on the autonomous intervention technology is a cross field,which requires the close cooperation between motion control,motion planning,navigation and visual perception.Motion planning of underwater vehicles is one of the most important part to realize autonomous intervention.In order to satisfy the sailing safety of underwater vehicle,the traditional motion planning technology of underwater vehicle is mainly to design a collision-free path from the start point to the destination,which includes geometric model-based method and potential field-based method,those methods have the disadvantages of low search efficiency and easy to fall into local convergence.Compared with the intervention requirements of traditional autonomous underwater vehicles,the intervention task and environment of sea creature capture robot is much more complicated,and the traditional motion planning can hardly meet the requirements of the intervention task.Therefore,in addition to decompose the task according to its purpose,more intelligent and advanced methods should be applied to the motion planning of the robot.The main research contents of this article are as follow.1)For the global research task of robot intervention in the marine ranching,a new global search planning using deep reinforcement learning is proposed.In order to increase the working efficiency of the robot in the marine ranching,the global search planning is decomposed.In order to satisfy flexible planning requirement,value iteration network is used for path planning of the robot,which guarantees the accuracy of the planning while improves the planning accuracy.In order to the optimization efficiency of the path planning,a particle swarm algorithm is used to optimize search order of the robot.Experiments prove that the global planning algorithm proposed in this paper can search the optimal path better and faster.2)For the local obstacle avoidance planning problem that the robot may encounter in marine ranching,an improved rapidly-exploring random tree algorithm(RRT)is proposed to plan the obstacle avoidance path.The improved RRT algorithm improves the shortcomings such as slow search efficiency and unsmooth paths,which has achieved obvious results.3)For the local capture task after the targets are found,a research on robot local capture planning using an improved multi-objective evolutionary algorithm is proposed.In order to meet the real-time requirements of robot online planning,a bare bone particle swarm optimization algorithm is used as the core algorithm of the evolutionary algorithm.In order to satisfy the search performance of the algorithm under multiple constraints,time-varying IFN thresholds and time-varying mutation strategy are used to improve search performance of this algorithm.Simulation experiments show that the improved multi-objective evolution algorithm can meet the needs of robot real-time online capture planning.