Real-time obstacle avoidance path planning based on bio-inspired neural network for unmanned surface vehicles
颜明重 1李琛 1朱大奇2
扫码查看
点击上方二维码区域，可以放大扫码查看
作者信息
1. 上海海事大学物流工程学院,上海 201306
2. 上海理工大学机械工程学院,上海 200093
折叠
摘要
针对无人水面艇(unmanned surface vehicle,USV)自主航行过程中的避障与遵守海事交通规则之间潜在的冲突问题,设计基于生物启发神经网络并且遵守《1972年国际海上避碰规则》(Convention on the International Regulations for Preventing Collisions At Sea,1972,COLREGs)的实时避障路径规划方法.运用STM32嵌入式平台搭建包括超声波、红外激光、陀螺仪和GPS传感器的小型USV水面环境感知硬件架构,将多传感器输出的动态环境信息通过栅格地图映射到二维神经网络中.USV根据神经网络活性势图自动规划通向目标点的无碰撞路径.通过多种船舶航行交汇局面的实验,证明该方法既安全又符合COLREGs的要求.
Abstract
To address potential conflicts between obstacle avoidance and maritime traffic rules during the autonomous navigation of unmanned surface vehicles(USVs),a real-time obstacle avoidance path planning method,which is based on a bio-inspired neural network and complies with the Convention on the International Regulations for Preventing Collisions at Sea,1972(COLREGs),is developed.STM32 embedded platform is used to build a small USV surface environment sensing hardware architecture including ultrasound,infrared laser,gyroscope,and GPS sensors.The dynamic environmental information output by multiple sensors is mapped to a two-dimensional neural network through a grid map.The USV automatically plans a collision-free path to the target based on the neural network activity potential diagram.The experiments on various ship navigation intersection situtations prove that this method is safe and meets the requirements of COLREGs.
维普
万方数据