AUV Online Path Planning In Three Dimensional Complex Underwater Space

When the AUV performs a task underwater,efficient path planning is the key to completing the task successfully.The multi-AUV system is a general form of underwater mission execution.This paper mainly studies the multi-AUV system to generate optimal allocation strategy and efficient path in real time and reduce the energy consumption through path planning in the process of executing tasks.The main work is as follows:(1)Aim at solving the problem of generating efficient multi-task execution paths in real time.An Reward acting on Reinforcement Learning and Particle Swarm Optimization(R-RLPSO)is proposed with real-time underwater multi-task assignment.This algorithm proposes a strategy of real-time rescue assignment based on rewards for the shortcomings of multi-AUV system that can not meet the real-time performance of underwater rescue assignment tasks.The multi-AUV system should perform rescue missions in different locations,each rescue mission is abstracted into a sphere rescue area,and propose the concept of attraction rescue area.A linear reward function is proposed to calculate the reward value for the path points that attracts the rescue area.In order to speed up the convergence of the R-RLPSO algorithm,a weighting factor is proposed to enhance the reward value of the rescue areas and the attraction rescue areas.The simulation results show that the reward value will accumulate in the rescue area,which proves the effectiveness of the R-RLPSO algorithm.(2)Aim at reducing the energy consumption through path planning during the process of performing tasks.This paper proposes a Distance Evolution Nonlinear Particle Swarm Optimization(DENPSO)to guide AUV to find an energy-efficient stable path.The innovation of the algorithm is to propose the distance evolution factor and define the evolution state.According to the evolution state,to avoid particles falling into local optimum regions,the particles of the poor search regions are randomly perturbed by the distance evolution state.A novel nonlinear inertia weighting factor and nonlinear learning factor are proposed,and the degree of nonlinearity is adaptively adjusted by defining hyperparameters.Each path is divided into several micro-element points based on the cubic spline interpolation method,the degree value factor and the micro-element collision factor are constructed by micro-element.The experimental results show that the energy consumption of the DENPSO algorithm is reduced by 2.1514e+03J and 1.049e+07J in the three-dimensional environment and the regional ocean model system.