仿生水下推进器的神经网络控制方法研究

项目名称： 仿生水下推进器的神经网络控制方法研究

项目编号： No.60805037

项目类型： 青年科学基金项目

立项/批准年度： 2009

项目学科： 生物科学

项目作者： 张代兵

作者单位： 中国人民解放军国防科学技术大学

项目金额： 16万元

中文摘要： 本项目围绕仿生水下推进器仿生神经网络控制方法开展研究，完成了鱼类生物推进器观测、仿生神经网络控制系统设计、仿生水下推进器实验验证等研究工作。首先，设计研制了双视角和三维同步成像实验观测系统，选取波动鳍方式推进的弓鳍目鱼类"尼罗河魔鬼"和身体尾鳍方式（BCF）推进的淡水大白鲨作为实验样本，结合国内外研究成果，开展了运动学观测、肌电信号测量和神经生理学实验与分析，对鱼类生物推进器的形态特征、动力学特性、控制特性及其中枢神经控制系统控制机理进行了研究。采用计算流体力学方法进行了仿生水下推进器推进的流体力学仿真实验和动力学分析。其次，设计完成了一种神经元数量少、动力学特性优异的新型人工神经元振荡器，以该振荡器为核心单元构建完成了模仿生物中枢模式发生器群（CPGs）机理的仿生神经网络控制系统，设计实现了仿生水下推进器各种运动状态的控制策略。最后，研制完成了仿生水下推进器试验装置，对控制系统的有效性和控制性能进行了试验验证。研究结果表明：该方法能够实现仿生水下推进器的灵活、高效的起动、巡航、停止等运动状态，使得不但从外形特征而且从运动状态、控制特性都更加接近，初步实现了"形似"到"神似"的进步。

中文关键词： 仿生；水下推进器；神经网络；中枢模式发生器群；控制系统

英文摘要： This project has investigated the bionic neural networks control method for bionic underwater propulsors，which consistes of fish biologic propulsor survey experiments,bionic neural networks control system design, bionic underwater propulsor tester experiment. Firstly, two fish locomotion synchronizational imaging observation system have been designed for observe in laboratory's water-tank. The typical Amiiform fish named " Gymnarchus niloticus" and Body-Caudal fin fish named "freshwater white shark" have been chosen as in observation experiments.The biologic propulsor's shape, distribution and skelecton muscle structure ，electromyogram experiments and analysis have been accomplished to explore the control principles in the fish's high performance CPGs neural control system. The dynamic behavior of undulating fins and underwater robot also been studied with computational fluid dynamic(CFD) method. Then, a new neural oscillator model which consists two neurons has been designed for simply dynamic properties. The bionic neural networks control system assembles many new neural ocscillators with inter-segmental connections. And the high level control rules and signals of different locomotion modes has be designed with the results of the dynamic analysis. Lastly, the control system prototype, the bionic propulsor tester and the hydrodynamic experiments on primary modes such as steady swimming, turning, starting and stop modevalidated the bionic neural network control method's feasible and hign performance.The advancement from the "Similarity in outline" of reverse kinematics control method to the "Similarity in spirit" of bionic neural network control method has been realized in this project.

英文关键词： bionic; underwater propulsor; neural network; central Pattern generators; control