Using inverse learning for controlling bionic robotic fish with SMA actuators
Using inverse learning for controlling bionic robotic fish with SMA actuators
复制标题
基于逆学习的形状记忆合金仿生机器鱼控制
DOI：
10.1557/s43580-022-00328-w
复制
发表时间：
2022-08
期刊：
影响因子：
0.8
通讯作者：
Kewei Ning;P. Hartono;H. Sawada
中科院分区：
文献类型：
--
作者：
Kewei Ning;P. Hartono;H. Sawada
在这项研究中，我们开发了一种用于游泳运动的无绳仿生软机器鱼。鱼的身体是用柔软的硅橡胶模塑的，我们使用形状记忆合金丝作为它的致动器。它的轻巧和灵活性使机器鱼能够产生仿生游泳动作。由于对机器人的游泳动力学进行数学建模的复杂性，建立一个逼真的模拟器是非常困难的。因此，在本研究中，我们引入前馈神经网络的逆学习来产生控制参数，以实现期望的游泳动作，并随后利用神经网络进行实时控制。本文介绍了我们研制的机器鱼的机电结构和神经控制器的实验。图形摘要
In this study, we develop an untethered bionic soft robotic fish for swimming motion. The body of the fish is molded using soft silicone rubber, and we utilize shape memory alloy wires for its actuators. Its lightness and flexibility allow the robotic fish to generate biomimetic swimming motions. Due to the complexity of mathematically modeling the robot’s swimming dynamics, building a realistic simulator is prohibitively difficult. Hence, in this study, we introduce inverse learning for a feedforward neural network to generate control parameters for realizing desired swimming motions and subsequently utilize the neural network for real-time control. In this paper, we report on the electro-mechanical structure of our robotic fish and the experiment of the neuro-controller. Graphical abstract