Unmanned ship path planning method based on deep reinforcement learning and considering marine environment elements

Abstract

The invention discloses an unmanned ship path planning method based on deep reinforcement learning and considering marine environment elements. The method comprises the basic steps of S1, interpolating wind, wave and flow data of a target sea area, and adding obstacle information, starting point information and end point information, S2, evaluating the maximum value of the wind wave flow borne bythe unmanned ship by using a Bayesian network, S3, reorganizing the AIS data of the target sea area into a training network to obtain an optimized experience pool and preliminary network parameters, S4, respectively inputting the unmanned ship state feature vectors into a deep reinforcement learning module for algorithm iteration, updating network parameters, and outputting actions,S5, iterating the unmanned ship to run for 15 seconds each time, and updating the data when the cumulative time reaches 1 hour,and S6, when the unmanned ship arrives at the target point, ending iteration, and outputting a path. According to the method, the influence of the marine environment factors on the navigation of the unmanned ship is fully considered, the actual long-range navigation condition of the unmanned ship is better met, the environment factors and obstacle information can be considered at the same time under the severe sea condition of the unmanned ship, and high-quality safety is obtained.

Description

technical field [0001] This patent relates to the field of unmanned ship path planning, in particular to an unmanned ship path planning method based on deep reinforcement learning and taking into account marine environment elements. Background technique [0002] Relying on the development of artificial intelligence control technology, unmanned ships have made breakthroughs in many technical fields, and gradually entered people's field of vision, began to undertake tasks such as ocean exploration and data acquisition, and gradually developed into the maritime transportation industry. [0003] Currently published patents: CN109657863A, CN109726866A, and CN107289939A all propose better path planning methods in this field, but generally only consider the impact of obstacles on unmanned ships. And each unmanned ship has the limit value of wind and waves according to its own material, structure and draft. When a ship sails in the sea, avoiding dangerous marine environment element...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the problem that many unmanned ship path planning methods proposed so far do not consider the marine environment elements, the present invention provides an unmanned ship path planning method based on deep reinforcement learning and taking into account the marine environment elements, fully considering the real marine environment elements and maritime obstacles, combined with deep reinforcement learning methods to plan a safe and efficient driving path for unmanned ships

Images