水下机器人神经网络滤波技术的研究

A filtering method with neural network for underwater vehicle

滤波技术能提高水下机器人传感器信息的可靠性和精度 ,以保障水下机器人智能作业的顺利完成。将水下机器人的动力学特性隐式地分布在网络权值上 ,构造BP神经网络 ,进行水下机器人的运动状态预报 ,并将预报值与平滑后的实测数据相结合进行滤波 ,有效去除信号的噪声。

Filtering can improve the reliability and precision of the sensor data to ensure the operation of the underwater vehicle (UV). This paper constructs BP neural network containing the dynamics of the UV in the linkweights of neural networks to predict the state. Filtered data can be obtained from the predicted value and measured data. The simulation results show that this method is very effective.