摘要
针对多波束声纳体积大,成本高的局限,利用单波束声呐的探测波束依次旋转,依次获取自主式水下航行器(autonomous underwater vehicle,AUV)前方的左、中、右3个区域的障碍物距离信息.通过设计合适的环境障碍状态与有效的避障行为集合,并利用强化学习选择适合AUV自主避障的障碍状态-行为组合.仿真实验表明,根据单波束传感器提供的障碍物信息,通过强化学习获得的状态-动作组合,可以保证AUV躲避前方90°开角的障碍物,达到安全航行的要求.
On one hand, the single-beam sonar acquires the obstacle distance information, which includes three areas （left, center and right） in front of autonomous underwater vehicle by rotating its ranging beam, for the large volume and high cost limitations of the multi-beam sonar.On the other hand,appropriate environmental states and effective obstacles avoidance behaviors are designed,and the proper state-action combinations for obstacle avoidance are selected by the reinforcement learning.Simulation results show that, according to the obstacle information provided by the single-beam sonar and the state-action combination obtained through reinforcement learning, AUV can guarantee to avoid obstacles in front of the opening angle of 90 degrees and meet requirements safe navigation.
出处
《厦门大学学报（自然科学版）》
CAS
CSCD
北大核心
2014年第4期484-489,共6页
Journal of Xiamen University：Natural Science
基金
国家自然科学基金(60975084,61165016)
关键词
自主式水下航行器
避障
单波束声呐
强化学习
autonomous underwater vehicle（AUV）
obstacle avoidance
single beam sonar
reinforcement learning
