摘要
为提高机器人姿态估计与感知精度,引进注意力机制,以微小型自主式水下机器人为例,设计一种全新的姿态估计方法。参照SCANet多分辨率感知网络架构,采用在其中插入信息补充、细节增强等功能模块的方式,建立微小型自主式水下机器人姿态多分辨率感知网络;将注意力机制划分为Ca-Block与Ca-Neck两个模块,提取微小型自主式水下机器人姿态中的特征信息;采用对信息融合处理的方式,设计机器人姿态信息归一化处理,以此实现微小型自主式水下机器人姿态自适应融合与估计。对比实验结果证明:设计的方法在实际应用中的效果良好,该方法可以提高姿态估计结果的精度,且姿态估计所需时间较短,仅需20s。因此,通过此种方式可以掌握机器人在水下的作业姿态,提高机器人工作时的工作效率。
In order to improve the attitude estimation and perception accuracy of the robot,attention mechanism is introduced,and a new attitude estimation method is designed,taking the micro autonomous underwater vehicle as an example.Referring to the SCANet multi-resolution perception network architecture,the micro autonomous underwater vehicle attitude multi-resolution perception network is established by inserting information supplement,detail enhancement and other functional modules;The attention mechanism is divided into two modules,Ca Block and Ca Neck,to extract the feature information in the attitude of the micro autonomous underwater vehicle;In order to realize the adaptive fusion and estimation of the micro autonomous underwater vehicle's attitude,the normalized processing of the robot's attitude information is designed by means of information fusion processing.The comparative experimental results show that the designed method has a good effect in practical application.This method can improve the accuracy of attitude estimation results,and the time required for attitude estimation is short,only 20 seconds.Therefore,through this way,we can master the underwater working posture of the robot and improve the working efficiency of the robot.
作者
边春华
张维
李邱达
文杰
Bian Chunhua;Zhang Wei;Li Qiuda;Wen Jie(China Nuclear Power Operations Management Co.,Zhejiang,Jiaxing,314300,China)
出处
《仪器仪表用户》
2023年第4期76-79,共4页
Instrumentation
关键词
注意力机制
估计方法
姿态
水下机器人
自主式
微小型
attention mechanism
estimation method
attitude
underwater robot
autonomous
micro type
