基于RBF-MPC的水下机器人避碰控制  

Collision Avoidance Control of Underwater Vehicle Based on RBF-MPC

作　　者：李晨 代笠 LI Chen;DAI Li(Military Representative Bureau of the Naval Armaments Department in Yichang,Yichang 430000,China;Military Representative Bureau of the Naval Armaments Department in Guangzhou,Guangzhou 510000,China)

机构地区：[1]海军装备部驻宜昌地区军事代表局,湖北宜昌443003 [2]海军装备部驻广州地区军事代表局,广东广州510000

出　　处：《数字海洋与水下攻防》2022年第5期443-447,共5页Digital Ocean & Underwater Warfare

摘　　要：水下机器人避碰控制是自主作业的重要基础,但复杂的约束条件和模型的不精确性增加了避障路径跟踪的技术难度。在传统模型预测控制的基础上,结合作业场景多种约束条件,引入径向基函数神经网络,提出了一种水平面避碰控制方法。首先,采用径向基神经网络建立误差补偿函数,提高传统动力学预测模型精度;然后,结合避碰路径跟踪控制,在滚动优化环节选取性能指标函数,并显式引入障碍物、执行机构与控制稳定性等约束条件;最后,通过仿真试验证明该方法能够控制水下机器人跟踪避碰路径实现水平面内障碍物规避。The collision avoidance control of underwater vehicles is an important basis for autonomous operation,but complicated constraints and imprecision of models increase the technical difficulty of path tracking in avoidance.Based on the predictive control of traditional model,this paper proposes a horizontal-plane collision avoidance control method,combining with various constraints of operation scene and introducing radial basis function neural network.Firstly,radial basis neural network is used to establish error compensation function to improve the accuracy of traditional dynamic predictive model.Then,combining with the tracking control of collision avoidance path,the performance index function is selected in the rolling horizon optimization stage,and the constraints such as obstacles,actuator and control stability are explicitly introduced.Finally,the simulation results show that the proposed method can control the underwater vehicle to track the collision avoidance path to achieve collision avoidance in the horizontal plane.

关 键 词：水下机器人 避碰控制 神经网络 模型预测控制 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]