水下机器人的S型模糊神经网络控制

S Model Fuzzy Neural Network Control of Underwater Vehicles

模糊神经网络控制已经成功应用于水下机器人运动控制中,但其运算过程和训练算法比较复杂,对嵌入式硬件要求也较高。根据带翼水下机器人的运动特性提出了S型模糊神经网络控制方法,并推导了网络权值学习算法,最后以XX水下机器人为研究对象进行了仿真实验。试验结果表明,与基于高斯型隶属函数的模糊神经网络控制器相比,在没有过多损失整体控制品质的情况下,其网络算法得到极大简化,运算速度得到了提高,反应能力增强,非常适用于对精确定位能力和运动速度要求不高,但要求高机动性的水下机器人。

Gauss membership function-based fuzzy neural network （FNN） is proved to be effective in motion control of underwater vehicles. However, its operational process and the training algorithm are complicated, placing great demands on embedded hardware. S model FNN was proposed according to the moving characters of underwater vehicles with wings. The learning algorithm was developed. The simulation results show that the modified FNN has simpler algorithm, higher calculation speed and improved response, compared with Gauss membership function-based FNN. It is applicable for the underwater vehicle which doesn＇t need to have the ability of accurate positioning but must have good maneuverability.